I want to find some causal relationship from some cols to label what i defined by from_pandas(no tears),but i got some causal relationship from label to others #63

ziyuwzf · 2020-07-30T07:13:13Z

Description

I want to find some causal relationship from some cols to label what i defined by from_pandas(no tears),but i got some causal relationship from label to others

Context

First,the dataset i have is 300K rows* 150cols.I defined the label(result) what i need to find what cols(reasons) contribute to,and the distribution of labels is more than 280,000 0 and more than 10,000 are 1.
Second,i use from_pandas() to learns the structure.
at the last,i use sm.remove_edges_below_threshold(0.01),but i got all causal relationship of label is that label contribute to others.

The text was updated successfully, but these errors were encountered:

* NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (mckinsey#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Update release.md * Address coments * Corrected release.md * fc1 to adjacency Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]>

* Merge back to develop * Simplifying viz.draw syntax in tutorial notebook (#46) * Add non negativity constraint in numpy lasso (#41) * Add plotting tutorial to the documentation (#47) * Unpin some requirements * Mixed type data generation (#55) Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Merge back to develop (#59) * Pytorch NOTEARS (#63) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Update release.md * Address coments * Corrected release.md * fc1 to adjacency Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> * Pinned sphinx-auto-doc-typehints (#66) * Corrected a spelling/grammar mistake (#55) * Fix/lint (#73) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * black * pin pytorch version * pin pytorch version Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Structure learning regressor (#68) * initial commit (local copy-paste) * fixed minor comments * minor bugfix * impute from children inital commit * bugfixes and method option * auto thresholding * autothreshold and bugfix * make threshold removal explicit * add l1 argument * remove child imputation * feat importance fix and tabu logic * moved threshold till dag * restructure with base class * coef mask * recipe * enable bias fitting * persist bias as node attribute * allow fit_intercept * minor PR comment fixes * minor comment adjustment * test coverage and l1 clarification * recipe * minor test fixes * more tests * full test coverage * revove python 3.5/3.6 unsupported import * add normalization option * idiomatic typing * correct pylint errors * update some tests * more typeing updates * more pylint requirements * more pylint disable * python 3.5 support * try to get to work with 3.5 * full coverage and 3.5 support * remove base class to pass test * remove unneeded supression * black formatting changes * remove unused import * pytlint supression * minor reformat change * isort fix * better defensive programming * fix unittests * docstring update * do Raises docstring properly * action SWE suggestions * hotfixes * minor update * minor black formatting change * final merge checkbox * fix end of file * Data Gen root node initialisation fix (#72) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * fix for consinuous normal data * generalise across all dtypes * support fit_intercept * fixed many test errors * test logic fixes * lint test fixes * python 3.5 failure change * minor test bugfix * black * pin pytorch version * pin pytorch version * additional test parameter * black formatting * requested changes * test updates and docstring * black format change * disable too many lines * change * move recipe to tutorial folder * releaseMD changes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: qbphilip <[email protected]> * [1/2] Poisson data for data gen (#61) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * variable name change bugfix * fix tests Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> * [2/2] Nonlinear Data gen (#60) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * add nonlinear * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * add nonlinear to init * change order in all * change release.md * root node fix on core + count * nonlinear support to wrappers * docstring update * bugfix and reproducability fix * many tests and test updates * poiss bugfix and test fix * moar test coverage * categorical dataframe test coverage * full test coverage and linting * fix linting and fstring * black reformat * fix unused pylint argument * pytest fix * FINAL linting fix * Fix stuff (#75) CircleCI fixes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Zain Patel <[email protected]> * update black version (#76) * fix black * Fix/check for NA or Infinity when notears is used (#54) * update scipy version (#77) * add DYNOTEARS implementation (#50) Adds DYNOTEARS and corresponding data generator (for testing) * Pytorch NOTEARS extension - Non-Linear/Hidden Layer (#65) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Add hidden layer * small refactor * directional adj * minor edits * fix bias issues * breaking changes update to the interface * typo * new regressor regularisation interface * update forward method * forward(X) predictions work * working! * bugfix data normalisation * some fixes * average regularisation and adj calc at end * give credit! Co-authored-by: Philip Pilgerstorfer <[email protected]> * loc lin docstring update Co-authored-by: Philip Pilgerstorfer <[email protected]> * docstring + fc1/fc2 name updates * moar docstring updates * more minor updates * remove normalize option * plotting util * rename to DAGRegressor * rename and checks * more util functions * fix bias * fix bias with no intercept * fix linear adj * add tests * minor fix * minor fixes * extend interface to bias * differentialte coef_ and feature_imporances * seperate bias element * tests * more test coverage * nonlinear test coverage * test hotfix * more test coverage * test requirements update * more test coverage * formatting changes * final pylint change * more linting * more bestpractice structuring * more minor fixes * FINAL linting updates * actual last change * update to reg defaults, additions to the tutorial * nonlinear regularisation updates * regressor tutorial * almost finishing touches * gradient based h function! * soft clamp and coef feature importance seperation * small api update, closer to batchnorm * docstring updates * stronger soft clamping * gradient L1 rather than L2 * fcpos neg removal, gradient optim * revert back to create_graph=True for 2nd derivative * remove print and test fix * black reformatting * new black version * full test coverage * isort fix * pylint fix * first layer h(W) for speed optimization * fix batch norm system * add nonlinear test * test hotfix * black reformat * isort fix * remove X requirement from h_func * regressor tutorial final commit and black update * LayerNorm replacement Co-authored-by: Philip Pilgerstorfer <[email protected]> * major changes * add standardization * minort changes * fix tests * rename reg parameters * linting * test coverage, docstting * check array for infs * fix isinstance to base type * fix isort, add test coverage * new tutorial * docstring fix Co-authored-by: Zain Patel <[email protected]> * test string match Co-authored-by: Zain Patel <[email protected]> * assert improvement Co-authored-by: Zain Patel <[email protected]> * SWE suggestions * minor bugfix * more test fixing Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * release.md, version bump, docs Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Jebq <[email protected]>

* Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * Release/0.8.0 (#80) * Merge back to develop * Simplifying viz.draw syntax in tutorial notebook (#46) * Add non negativity constraint in numpy lasso (#41) * Add plotting tutorial to the documentation (#47) * Unpin some requirements * Mixed type data generation (#55) Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Merge back to develop (#59) * Pytorch NOTEARS (#63) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Update release.md * Address coments * Corrected release.md * fc1 to adjacency Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> * Pinned sphinx-auto-doc-typehints (#66) * Corrected a spelling/grammar mistake (#55) * Fix/lint (#73) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * black * pin pytorch version * pin pytorch version Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Structure learning regressor (#68) * initial commit (local copy-paste) * fixed minor comments * minor bugfix * impute from children inital commit * bugfixes and method option * auto thresholding * autothreshold and bugfix * make threshold removal explicit * add l1 argument * remove child imputation * feat importance fix and tabu logic * moved threshold till dag * restructure with base class * coef mask * recipe * enable bias fitting * persist bias as node attribute * allow fit_intercept * minor PR comment fixes * minor comment adjustment * test coverage and l1 clarification * recipe * minor test fixes * more tests * full test coverage * revove python 3.5/3.6 unsupported import * add normalization option * idiomatic typing * correct pylint errors * update some tests * more typeing updates * more pylint requirements * more pylint disable * python 3.5 support * try to get to work with 3.5 * full coverage and 3.5 support * remove base class to pass test * remove unneeded supression * black formatting changes * remove unused import * pytlint supression * minor reformat change * isort fix * better defensive programming * fix unittests * docstring update * do Raises docstring properly * action SWE suggestions * hotfixes * minor update * minor black formatting change * final merge checkbox * fix end of file * Data Gen root node initialisation fix (#72) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * fix for consinuous normal data * generalise across all dtypes * support fit_intercept * fixed many test errors * test logic fixes * lint test fixes * python 3.5 failure change * minor test bugfix * black * pin pytorch version * pin pytorch version * additional test parameter * black formatting * requested changes * test updates and docstring * black format change * disable too many lines * change * move recipe to tutorial folder * releaseMD changes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: qbphilip <[email protected]> * [1/2] Poisson data for data gen (#61) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * variable name change bugfix * fix tests Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> * [2/2] Nonlinear Data gen (#60) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * add nonlinear * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * add nonlinear to init * change order in all * change release.md * root node fix on core + count * nonlinear support to wrappers * docstring update * bugfix and reproducability fix * many tests and test updates * poiss bugfix and test fix * moar test coverage * categorical dataframe test coverage * full test coverage and linting * fix linting and fstring * black reformat * fix unused pylint argument * pytest fix * FINAL linting fix * Fix stuff (#75) CircleCI fixes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Zain Patel <[email protected]> * update black version (#76) * fix black * Fix/check for NA or Infinity when notears is used (#54) * update scipy version (#77) * add DYNOTEARS implementation (#50) Adds DYNOTEARS and corresponding data generator (for testing) * Pytorch NOTEARS extension - Non-Linear/Hidden Layer (#65) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Add hidden layer * small refactor * directional adj * minor edits * fix bias issues * breaking changes update to the interface * typo * new regressor regularisation interface * update forward method * forward(X) predictions work * working! * bugfix data normalisation * some fixes * average regularisation and adj calc at end * give credit! Co-authored-by: Philip Pilgerstorfer <[email protected]> * loc lin docstring update Co-authored-by: Philip Pilgerstorfer <[email protected]> * docstring + fc1/fc2 name updates * moar docstring updates * more minor updates * remove normalize option * plotting util * rename to DAGRegressor * rename and checks * more util functions * fix bias * fix bias with no intercept * fix linear adj * add tests * minor fix * minor fixes * extend interface to bias * differentialte coef_ and feature_imporances * seperate bias element * tests * more test coverage * nonlinear test coverage * test hotfix * more test coverage * test requirements update * more test coverage * formatting changes * final pylint change * more linting * more bestpractice structuring * more minor fixes * FINAL linting updates * actual last change * update to reg defaults, additions to the tutorial * nonlinear regularisation updates * regressor tutorial * almost finishing touches * gradient based h function! * soft clamp and coef feature importance seperation * small api update, closer to batchnorm * docstring updates * stronger soft clamping * gradient L1 rather than L2 * fcpos neg removal, gradient optim * revert back to create_graph=True for 2nd derivative * remove print and test fix * black reformatting * new black version * full test coverage * isort fix * pylint fix * first layer h(W) for speed optimization * fix batch norm system * add nonlinear test * test hotfix * black reformat * isort fix * remove X requirement from h_func * regressor tutorial final commit and black update * LayerNorm replacement Co-authored-by: Philip Pilgerstorfer <[email protected]> * major changes * add standardization * minort changes * fix tests * rename reg parameters * linting * test coverage, docstting * check array for infs * fix isinstance to base type * fix isort, add test coverage * new tutorial * docstring fix Co-authored-by: Zain Patel <[email protected]> * test string match Co-authored-by: Zain Patel <[email protected]> * assert improvement Co-authored-by: Zain Patel <[email protected]> * SWE suggestions * minor bugfix * more test fixing Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * release.md, version bump, docs Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Jebq <[email protected]> * release.MD * version bump * Update causalnex/structure/pytorch/dist_type/_base.py Co-authored-by: Zain Patel <[email protected]> * Update causalnex/structure/pytorch/dist_type/__init__.py Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Jebq <[email protected]>

* Merge back to develop * Simplifying viz.draw syntax in tutorial notebook (#46) * Add non negativity constraint in numpy lasso (#41) * Add plotting tutorial to the documentation (#47) * Unpin some requirements * Mixed type data generation (#55) Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Merge back to develop (#59) * Pytorch NOTEARS (#63) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Update release.md * Address coments * Corrected release.md * fc1 to adjacency Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> * Pinned sphinx-auto-doc-typehints (#66) * Corrected a spelling/grammar mistake (#55) * Fix/lint (#73) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * black * pin pytorch version * pin pytorch version Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Structure learning regressor (#68) * initial commit (local copy-paste) * fixed minor comments * minor bugfix * impute from children inital commit * bugfixes and method option * auto thresholding * autothreshold and bugfix * make threshold removal explicit * add l1 argument * remove child imputation * feat importance fix and tabu logic * moved threshold till dag * restructure with base class * coef mask * recipe * enable bias fitting * persist bias as node attribute * allow fit_intercept * minor PR comment fixes * minor comment adjustment * test coverage and l1 clarification * recipe * minor test fixes * more tests * full test coverage * revove python 3.5/3.6 unsupported import * add normalization option * idiomatic typing * correct pylint errors * update some tests * more typeing updates * more pylint requirements * more pylint disable * python 3.5 support * try to get to work with 3.5 * full coverage and 3.5 support * remove base class to pass test * remove unneeded supression * black formatting changes * remove unused import * pytlint supression * minor reformat change * isort fix * better defensive programming * fix unittests * docstring update * do Raises docstring properly * action SWE suggestions * hotfixes * minor update * minor black formatting change * final merge checkbox * fix end of file * Data Gen root node initialisation fix (#72) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * fix for consinuous normal data * generalise across all dtypes * support fit_intercept * fixed many test errors * test logic fixes * lint test fixes * python 3.5 failure change * minor test bugfix * black * pin pytorch version * pin pytorch version * additional test parameter * black formatting * requested changes * test updates and docstring * black format change * disable too many lines * change * move recipe to tutorial folder * releaseMD changes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: qbphilip <[email protected]> * [1/2] Poisson data for data gen (#61) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * variable name change bugfix * fix tests Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> * [2/2] Nonlinear Data gen (#60) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * add nonlinear * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * add nonlinear to init * change order in all * change release.md * root node fix on core + count * nonlinear support to wrappers * docstring update * bugfix and reproducability fix * many tests and test updates * poiss bugfix and test fix * moar test coverage * categorical dataframe test coverage * full test coverage and linting * fix linting and fstring * black reformat * fix unused pylint argument * pytest fix * FINAL linting fix * Fix stuff (#75) CircleCI fixes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Zain Patel <[email protected]> * update black version (#76) * fix black * Fix/check for NA or Infinity when notears is used (#54) * update scipy version (#77) * add DYNOTEARS implementation (#50) Adds DYNOTEARS and corresponding data generator (for testing) * Pytorch NOTEARS extension - Non-Linear/Hidden Layer (#65) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Add hidden layer * small refactor * directional adj * minor edits * fix bias issues * breaking changes update to the interface * typo * new regressor regularisation interface * update forward method * forward(X) predictions work * working! * bugfix data normalisation * some fixes * average regularisation and adj calc at end * give credit! Co-authored-by: Philip Pilgerstorfer <[email protected]> * loc lin docstring update Co-authored-by: Philip Pilgerstorfer <[email protected]> * docstring + fc1/fc2 name updates * moar docstring updates * more minor updates * remove normalize option * plotting util * rename to DAGRegressor * rename and checks * more util functions * fix bias * fix bias with no intercept * fix linear adj * add tests * minor fix * minor fixes * extend interface to bias * differentialte coef_ and feature_imporances * seperate bias element * tests * more test coverage * nonlinear test coverage * test hotfix * more test coverage * test requirements update * more test coverage * formatting changes * final pylint change * more linting * more bestpractice structuring * more minor fixes * FINAL linting updates * actual last change * update to reg defaults, additions to the tutorial * nonlinear regularisation updates * regressor tutorial * almost finishing touches * gradient based h function! * soft clamp and coef feature importance seperation * small api update, closer to batchnorm * docstring updates * stronger soft clamping * gradient L1 rather than L2 * fcpos neg removal, gradient optim * revert back to create_graph=True for 2nd derivative * remove print and test fix * black reformatting * new black version * full test coverage * isort fix * pylint fix * first layer h(W) for speed optimization * fix batch norm system * add nonlinear test * test hotfix * black reformat * isort fix * remove X requirement from h_func * regressor tutorial final commit and black update * LayerNorm replacement Co-authored-by: Philip Pilgerstorfer <[email protected]> * major changes * add standardization * minort changes * fix tests * rename reg parameters * linting * test coverage, docstting * check array for infs * fix isinstance to base type * fix isort, add test coverage * new tutorial * docstring fix Co-authored-by: Zain Patel <[email protected]> * test string match Co-authored-by: Zain Patel <[email protected]> * assert improvement Co-authored-by: Zain Patel <[email protected]> * SWE suggestions * minor bugfix * more test fixing Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * Merge release/0.8.0 back into develop for release 0.8.0 (#82) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * release.md, version bump, docs Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * fix tests (#87) * Fix pygments fail (#84) Thanks Zain! * update notebook beginning (#89) * Add Binary distribution type support (#85) * binary dtype folder and __init__ * dtype base class * continuous dtype * binary dtype * update core * make plural * update interface for idx * minor variable name change * notears update * python 3.5 support * fix fstring * remove categorical methods, doctrings * formatting and docstrings * remove redundant cat code * isort * indexerror check * defensive check tests * datatype loss tests * more test coverage * more tests and formatting * fix test import * remove double test * linting * docstring and pylint * docstring fix Co-authored-by: Zain Patel <[email protected]> * fix long string Co-authored-by: Zain Patel <[email protected]> * docstring fix Co-authored-by: Zain Patel <[email protected]> * remove relative imports * docstring fix * dict comprehension * list comprehension and neatness * remove unuesd import to __init__ * fix test * remove unused return interface * add binary f1score tests * one datatype instane per feature * rename dtype -> disttype, attach dists to nodes * fix tests * fix linting * fix preserve node dtyper * fix tests * fix tests * fix tests * final docstring and test fixes * lint fix * test_fix, warning * linting * fix test * fix tests * reduce threshold of test * raise better error * black linting * remove warning * remove useless supression and import Co-authored-by: Zain Patel <[email protected]> * Add sklearn binary classifier (#90) * binary dtype folder and __init__ * dtype base class * continuous dtype * binary dtype * update core * make plural * update interface for idx * minor variable name change * notears update * python 3.5 support * fix fstring * remove categorical methods, doctrings * formatting and docstrings * remove redundant cat code * isort * indexerror check * defensive check tests * datatype loss tests * more test coverage * more tests and formatting * fix test import * remove double test * linting * docstring and pylint * docstring fix Co-authored-by: Zain Patel <[email protected]> * fix long string Co-authored-by: Zain Patel <[email protected]> * docstring fix Co-authored-by: Zain Patel <[email protected]> * remove relative imports * docstring fix * dict comprehension * list comprehension and neatness * remove unuesd import to __init__ * fix test * remove unused return interface * new sklearn folder structure * sklearn class outline * new dtype interface * docstring clarification * inverse link function * add binary f1score tests * one datatype instane per feature * rename dtype -> disttype, attach dists to nodes * fix tests * fix linting * fix preserve node dtyper * fix tests * fix tests * fix tests * final docstring and test fixes * lint fix * test_fix, warning * linting * fix test * fix tests * reduce threshold of test * docstring clarification * _target_dist_type injection * docstring updates + clf fit outline * old doctring deprecation * clf predict_proba and predict * return bugfix * docstring update * import fix, linting, clf fit finished * args docstring and None schema * raise better error * black linting * linting * revert to public interface * remove warning * remove useless supression and import * add useless change to resolve merge conflict * update inits * standardization and data reconstruction * remove unused imports * fix clf .precit() * regressor fit_predict * remove useless regressor predict * test import fix * fix warnings * pass series name thru * fix schema pass thru * better dict comprehension Co-authored-by: Zain Patel <[email protected]> * import and comment fixes * update to .format() * fig sklearn is fitted test * more dtype schema insertion * DAGRegressor test fix * dag regressor test * more linting * big test restructure * combined test suite * error string update * more test coverage * linting and isort * move test to combined test * return float64 preds * moar clf tests * remove untestable (multiclass) code * class number error test * balck reformat * docstrings, pylint * fix test bug * standard scaler for _base * pull classes direct from LabelEncoder * update tutorial Co-authored-by: Zain Patel <[email protected]> * Fix/uniform discretisation (#65) * Fix uniform discretiser * Fix uniform discretiser Co-authored-by: Philip Pilgerstorfer <[email protected]> * release.MD * version bump * Update causalnex/structure/pytorch/dist_type/_base.py Co-authored-by: Zain Patel <[email protected]> * Update causalnex/structure/pytorch/dist_type/__init__.py Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Jebq <[email protected]> Co-authored-by: Shuhei Ishida <[email protected]>

* Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * Release/0.8.0 (#80) * Merge back to develop * Simplifying viz.draw syntax in tutorial notebook (#46) * Add non negativity constraint in numpy lasso (#41) * Add plotting tutorial to the documentation (#47) * Unpin some requirements * Mixed type data generation (#55) Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Merge back to develop (#59) * Pytorch NOTEARS (#63) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Update release.md * Address coments * Corrected release.md * fc1 to adjacency Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> * Pinned sphinx-auto-doc-typehints (#66) * Corrected a spelling/grammar mistake (#55) * Fix/lint (#73) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * black * pin pytorch version * pin pytorch version Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Structure learning regressor (#68) * initial commit (local copy-paste) * fixed minor comments * minor bugfix * impute from children inital commit * bugfixes and method option * auto thresholding * autothreshold and bugfix * make threshold removal explicit * add l1 argument * remove child imputation * feat importance fix and tabu logic * moved threshold till dag * restructure with base class * coef mask * recipe * enable bias fitting * persist bias as node attribute * allow fit_intercept * minor PR comment fixes * minor comment adjustment * test coverage and l1 clarification * recipe * minor test fixes * more tests * full test coverage * revove python 3.5/3.6 unsupported import * add normalization option * idiomatic typing * correct pylint errors * update some tests * more typeing updates * more pylint requirements * more pylint disable * python 3.5 support * try to get to work with 3.5 * full coverage and 3.5 support * remove base class to pass test * remove unneeded supression * black formatting changes * remove unused import * pytlint supression * minor reformat change * isort fix * better defensive programming * fix unittests * docstring update * do Raises docstring properly * action SWE suggestions * hotfixes * minor update * minor black formatting change * final merge checkbox * fix end of file * Data Gen root node initialisation fix (#72) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * fix for consinuous normal data * generalise across all dtypes * support fit_intercept * fixed many test errors * test logic fixes * lint test fixes * python 3.5 failure change * minor test bugfix * black * pin pytorch version * pin pytorch version * additional test parameter * black formatting * requested changes * test updates and docstring * black format change * disable too many lines * change * move recipe to tutorial folder * releaseMD changes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: qbphilip <[email protected]> * [1/2] Poisson data for data gen (#61) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * variable name change bugfix * fix tests Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> * [2/2] Nonlinear Data gen (#60) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * add nonlinear * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * add nonlinear to init * change order in all * change release.md * root node fix on core + count * nonlinear support to wrappers * docstring update * bugfix and reproducability fix * many tests and test updates * poiss bugfix and test fix * moar test coverage * categorical dataframe test coverage * full test coverage and linting * fix linting and fstring * black reformat * fix unused pylint argument * pytest fix * FINAL linting fix * Fix stuff (#75) CircleCI fixes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Zain Patel <[email protected]> * update black version (#76) * fix black * Fix/check for NA or Infinity when notears is used (#54) * update scipy version (#77) * add DYNOTEARS implementation (#50) Adds DYNOTEARS and corresponding data generator (for testing) * Pytorch NOTEARS extension - Non-Linear/Hidden Layer (#65) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Add hidden layer * small refactor * directional adj * minor edits * fix bias issues * breaking changes update to the interface * typo * new regressor regularisation interface * update forward method * forward(X) predictions work * working! * bugfix data normalisation * some fixes * average regularisation and adj calc at end * give credit! Co-authored-by: Philip Pilgerstorfer <[email protected]> * loc lin docstring update Co-authored-by: Philip Pilgerstorfer <[email protected]> * docstring + fc1/fc2 name updates * moar docstring updates * more minor updates * remove normalize option * plotting util * rename to DAGRegressor * rename and checks * more util functions * fix bias * fix bias with no intercept * fix linear adj * add tests * minor fix * minor fixes * extend interface to bias * differentialte coef_ and feature_imporances * seperate bias element * tests * more test coverage * nonlinear test coverage * test hotfix * more test coverage * test requirements update * more test coverage * formatting changes * final pylint change * more linting * more bestpractice structuring * more minor fixes * FINAL linting updates * actual last change * update to reg defaults, additions to the tutorial * nonlinear regularisation updates * regressor tutorial * almost finishing touches * gradient based h function! * soft clamp and coef feature importance seperation * small api update, closer to batchnorm * docstring updates * stronger soft clamping * gradient L1 rather than L2 * fcpos neg removal, gradient optim * revert back to create_graph=True for 2nd derivative * remove print and test fix * black reformatting * new black version * full test coverage * isort fix * pylint fix * first layer h(W) for speed optimization * fix batch norm system * add nonlinear test * test hotfix * black reformat * isort fix * remove X requirement from h_func * regressor tutorial final commit and black update * LayerNorm replacement Co-authored-by: Philip Pilgerstorfer <[email protected]> * major changes * add standardization * minort changes * fix tests * rename reg parameters * linting * test coverage, docstting * check array for infs * fix isinstance to base type * fix isort, add test coverage * new tutorial * docstring fix Co-authored-by: Zain Patel <[email protected]> * test string match Co-authored-by: Zain Patel <[email protected]> * assert improvement Co-authored-by: Zain Patel <[email protected]> * SWE suggestions * minor bugfix * more test fixing Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * release.md, version bump, docs Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Jebq <[email protected]> * Release/0.8.1 (#94) * Merge back to develop * Simplifying viz.draw syntax in tutorial notebook (#46) * Add non negativity constraint in numpy lasso (#41) * Add plotting tutorial to the documentation (#47) * Unpin some requirements * Mixed type data generation (#55) Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Merge back to develop (#59) * Pytorch NOTEARS (#63) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Update release.md * Address coments * Corrected release.md * fc1 to adjacency Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> * Pinned sphinx-auto-doc-typehints (#66) * Corrected a spelling/grammar mistake (#55) * Fix/lint (#73) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * black * pin pytorch version * pin pytorch version Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Structure learning regressor (#68) * initial commit (local copy-paste) * fixed minor comments * minor bugfix * impute from children inital commit * bugfixes and method option * auto thresholding * autothreshold and bugfix * make threshold removal explicit * add l1 argument * remove child imputation * feat importance fix and tabu logic * moved threshold till dag * restructure with base class * coef mask * recipe * enable bias fitting * persist bias as node attribute * allow fit_intercept * minor PR comment fixes * minor comment adjustment * test coverage and l1 clarification * recipe * minor test fixes * more tests * full test coverage * revove python 3.5/3.6 unsupported import * add normalization option * idiomatic typing * correct pylint errors * update some tests * more typeing updates * more pylint requirements * more pylint disable * python 3.5 support * try to get to work with 3.5 * full coverage and 3.5 support * remove base class to pass test * remove unneeded supression * black formatting changes * remove unused import * pytlint supression * minor reformat change * isort fix * better defensive programming * fix unittests * docstring update * do Raises docstring properly * action SWE suggestions * hotfixes * minor update * minor black formatting change * final merge checkbox * fix end of file * Data Gen root node initialisation fix (#72) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * fix for consinuous normal data * generalise across all dtypes * support fit_intercept * fixed many test errors * test logic fixes * lint test fixes * python 3.5 failure change * minor test bugfix * black * pin pytorch version * pin pytorch version * additional test parameter * black formatting * requested changes * test updates and docstring * black format change * disable too many lines * change * move recipe to tutorial folder * releaseMD changes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: qbphilip <[email protected]> * [1/2] Poisson data for data gen (#61) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * variable name change bugfix * fix tests Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> * [2/2] Nonlinear Data gen (#60) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * add nonlinear * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * add nonlinear to init * change order in all * change release.md * root node fix on core + count * nonlinear support to wrappers * docstring update * bugfix and reproducability fix * many tests and test updates * poiss bugfix and test fix * moar test coverage * categorical dataframe test coverage * full test coverage and linting * fix linting and fstring * black reformat * fix unused pylint argument * pytest fix * FINAL linting fix * Fix stuff (#75) CircleCI fixes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Zain Patel <[email protected]> * update black version (#76) * fix black * Fix/check for NA or Infinity when notears is used (#54) * update scipy version (#77) * add DYNOTEARS implementation (#50) Adds DYNOTEARS and corresponding data generator (for testing) * Pytorch NOTEARS extension - Non-Linear/Hidden Layer (#65) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Add hidden layer * small refactor * directional adj * minor edits * fix bias issues * breaking changes update to the interface * typo * new regressor regularisation interface * update forward method * forward(X) predictions work * working! * bugfix data normalisation * some fixes * average regularisation and adj calc at end * give credit! Co-authored-by: Philip Pilgerstorfer <[email protected]> * loc lin docstring update Co-authored-by: Philip Pilgerstorfer <[email protected]> * docstring + fc1/fc2 name updates * moar docstring updates * more minor updates * remove normalize option * plotting util * rename to DAGRegressor * rename and checks * more util functions * fix bias * fix bias with no intercept * fix linear adj * add tests * minor fix * minor fixes * extend interface to bias * differentialte coef_ and feature_imporances * seperate bias element * tests * more test coverage * nonlinear test coverage * test hotfix * more test coverage * test requirements update * more test coverage * formatting changes * final pylint change * more linting * more bestpractice structuring * more minor fixes * FINAL linting updates * actual last change * update to reg defaults, additions to the tutorial * nonlinear regularisation updates * regressor tutorial * almost finishing touches * gradient based h function! * soft clamp and coef feature importance seperation * small api update, closer to batchnorm * docstring updates * stronger soft clamping * gradient L1 rather than L2 * fcpos neg removal, gradient optim * revert back to create_graph=True for 2nd derivative * remove print and test fix * black reformatting * new black version * full test coverage * isort fix * pylint fix * first layer h(W) for speed optimization * fix batch norm system * add nonlinear test * test hotfix * black reformat * isort fix * remove X requirement from h_func * regressor tutorial final commit and black update * LayerNorm replacement Co-authored-by: Philip Pilgerstorfer <[email protected]> * major changes * add standardization * minort changes * fix tests * rename reg parameters * linting * test coverage, docstting * check array for infs * fix isinstance to base type * fix isort, add test coverage * new tutorial * docstring fix Co-authored-by: Zain Patel <[email protected]> * test string match Co-authored-by: Zain Patel <[email protected]> * assert improvement Co-authored-by: Zain Patel <[email protected]> * SWE suggestions * minor bugfix * more test fixing Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * Merge release/0.8.0 back into develop for release 0.8.0 (#82) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * release.md, version bump, docs Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * fix tests (#87) * Fix pygments fail (#84) Thanks Zain! * update notebook beginning (#89) * Add Binary distribution type support (#85) * binary dtype folder and __init__ * dtype base class * continuous dtype * binary dtype * update core * make plural * update interface for idx * minor variable name change * notears update * python 3.5 support * fix fstring * remove categorical methods, doctrings * formatting and docstrings * remove redundant cat code * isort * indexerror check * defensive check tests * datatype loss tests * more test coverage * more tests and formatting * fix test import * remove double test * linting * docstring and pylint * docstring fix Co-authored-by: Zain Patel <[email protected]> * fix long string Co-authored-by: Zain Patel <[email protected]> * docstring fix Co-authored-by: Zain Patel <[email protected]> * remove relative imports * docstring fix * dict comprehension * list comprehension and neatness * remove unuesd import to __init__ * fix test * remove unused return interface * add binary f1score tests * one datatype instane per feature * rename dtype -> disttype, attach dists to nodes * fix tests * fix linting * fix preserve node dtyper * fix tests * fix tests * fix tests * final docstring and test fixes * lint fix * test_fix, warning * linting * fix test * fix tests * reduce threshold of test * raise better error * black linting * remove warning * remove useless supression and import Co-authored-by: Zain Patel <[email protected]> * Add sklearn binary classifier (#90) * binary dtype folder and __init__ * dtype base class * continuous dtype * binary dtype * update core * make plural * update interface for idx * minor variable name change * notears update * python 3.5 support * fix fstring * remove categorical methods, doctrings * formatting and docstrings * remove redundant cat code * isort * indexerror check * defensive check tests * datatype loss tests * more test coverage * more tests and formatting * fix test import * remove double test * linting * docstring and pylint * docstring fix Co-authored-by: Zain Patel <[email protected]> * fix long string Co-authored-by: Zain Patel <[email protected]> * docstring fix Co-authored-by: Zain Patel <[email protected]> * remove relative imports * docstring fix * dict comprehension * list comprehension and neatness * remove unuesd import to __init__ * fix test * remove unused return interface * new sklearn folder structure * sklearn class outline * new dtype interface * docstring clarification * inverse link function * add binary f1score tests * one datatype instane per feature * rename dtype -> disttype, attach dists to nodes * fix tests * fix linting * fix preserve node dtyper * fix tests * fix tests * fix tests * final docstring and test fixes * lint fix * test_fix, warning * linting * fix test * fix tests * reduce threshold of test * docstring clarification * _target_dist_type injection * docstring updates + clf fit outline * old doctring deprecation * clf predict_proba and predict * return bugfix * docstring update * import fix, linting, clf fit finished * args docstring and None schema * raise better error * black linting * linting * revert to public interface * remove warning * remove useless supression and import * add useless change to resolve merge conflict * update inits * standardization and data reconstruction * remove unused imports * fix clf .precit() * regressor fit_predict * remove useless regressor predict * test import fix * fix warnings * pass series name thru * fix schema pass thru * better dict comprehension Co-authored-by: Zain Patel <[email protected]> * import and comment fixes * update to .format() * fig sklearn is fitted test * more dtype schema insertion * DAGRegressor test fix * dag regressor test * more linting * big test restructure * combined test suite * error string update * more test coverage * linting and isort * move test to combined test * return float64 preds * moar clf tests * remove untestable (multiclass) code * class number error test * balck reformat * docstrings, pylint * fix test bug * standard scaler for _base * pull classes direct from LabelEncoder * update tutorial Co-authored-by: Zain Patel <[email protected]> * Fix/uniform discretisation (#65) * Fix uniform discretiser * Fix uniform discretiser Co-authored-by: Philip Pilgerstorfer <[email protected]> * release.MD * version bump * Update causalnex/structure/pytorch/dist_type/_base.py Co-authored-by: Zain Patel <[email protected]> * Update causalnex/structure/pytorch/dist_type/__init__.py Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Jebq <[email protected]> Co-authored-by: Shuhei Ishida <[email protected]> * deprecate python3.5 * modify setup.py * fix * better black linting * fix tests * fix break coverage * Update RELEASE.md Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: Angel Droth <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Jebq <[email protected]> Co-authored-by: Shuhei Ishida <[email protected]> Co-authored-by: angeldrothqb <[email protected]>

* Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * Release/0.8.0 (#80) * Merge back to develop * Simplifying viz.draw syntax in tutorial notebook (#46) * Add non negativity constraint in numpy lasso (#41) * Add plotting tutorial to the documentation (#47) * Unpin some requirements * Mixed type data generation (#55) Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Merge back to develop (#59) * Pytorch NOTEARS (#63) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Update release.md * Address coments * Corrected release.md * fc1 to adjacency Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> * Pinned sphinx-auto-doc-typehints (#66) * Corrected a spelling/grammar mistake (#55) * Fix/lint (#73) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * black * pin pytorch version * pin pytorch version Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Structure learning regressor (#68) * initial commit (local copy-paste) * fixed minor comments * minor bugfix * impute from children inital commit * bugfixes and method option * auto thresholding * autothreshold and bugfix * make threshold removal explicit * add l1 argument * remove child imputation * feat importance fix and tabu logic * moved threshold till dag * restructure with base class * coef mask * recipe * enable bias fitting * persist bias as node attribute * allow fit_intercept * minor PR comment fixes * minor comment adjustment * test coverage and l1 clarification * recipe * minor test fixes * more tests * full test coverage * revove python 3.5/3.6 unsupported import * add normalization option * idiomatic typing * correct pylint errors * update some tests * more typeing updates * more pylint requirements * more pylint disable * python 3.5 support * try to get to work with 3.5 * full coverage and 3.5 support * remove base class to pass test * remove unneeded supression * black formatting changes * remove unused import * pytlint supression * minor reformat change * isort fix * better defensive programming * fix unittests * docstring update * do Raises docstring properly * action SWE suggestions * hotfixes * minor update * minor black formatting change * final merge checkbox * fix end of file * Data Gen root node initialisation fix (#72) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * fix for consinuous normal data * generalise across all dtypes * support fit_intercept * fixed many test errors * test logic fixes * lint test fixes * python 3.5 failure change * minor test bugfix * black * pin pytorch version * pin pytorch version * additional test parameter * black formatting * requested changes * test updates and docstring * black format change * disable too many lines * change * move recipe to tutorial folder * releaseMD changes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: qbphilip <[email protected]> * [1/2] Poisson data for data gen (#61) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * variable name change bugfix * fix tests Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> * [2/2] Nonlinear Data gen (#60) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * add nonlinear * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * add nonlinear to init * change order in all * change release.md * root node fix on core + count * nonlinear support to wrappers * docstring update * bugfix and reproducability fix * many tests and test updates * poiss bugfix and test fix * moar test coverage * categorical dataframe test coverage * full test coverage and linting * fix linting and fstring * black reformat * fix unused pylint argument * pytest fix * FINAL linting fix * Fix stuff (#75) CircleCI fixes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Zain Patel <[email protected]> * update black version (#76) * fix black * Fix/check for NA or Infinity when notears is used (#54) * update scipy version (#77) * add DYNOTEARS implementation (#50) Adds DYNOTEARS and corresponding data generator (for testing) * Pytorch NOTEARS extension - Non-Linear/Hidden Layer (#65) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Add hidden layer * small refactor * directional adj * minor edits * fix bias issues * breaking changes update to the interface * typo * new regressor regularisation interface * update forward method * forward(X) predictions work * working! * bugfix data normalisation * some fixes * average regularisation and adj calc at end * give credit! Co-authored-by: Philip Pilgerstorfer <[email protected]> * loc lin docstring update Co-authored-by: Philip Pilgerstorfer <[email protected]> * docstring + fc1/fc2 name updates * moar docstring updates * more minor updates * remove normalize option * plotting util * rename to DAGRegressor * rename and checks * more util functions * fix bias * fix bias with no intercept * fix linear adj * add tests * minor fix * minor fixes * extend interface to bias * differentialte coef_ and feature_imporances * seperate bias element * tests * more test coverage * nonlinear test coverage * test hotfix * more test coverage * test requirements update * more test coverage * formatting changes * final pylint change * more linting * more bestpractice structuring * more minor fixes * FINAL linting updates * actual last change * update to reg defaults, additions to the tutorial * nonlinear regularisation updates * regressor tutorial * almost finishing touches * gradient based h function! * soft clamp and coef feature importance seperation * small api update, closer to batchnorm * docstring updates * stronger soft clamping * gradient L1 rather than L2 * fcpos neg removal, gradient optim * revert back to create_graph=True for 2nd derivative * remove print and test fix * black reformatting * new black version * full test coverage * isort fix * pylint fix * first layer h(W) for speed optimization * fix batch norm system * add nonlinear test * test hotfix * black reformat * isort fix * remove X requirement from h_func * regressor tutorial final commit and black update * LayerNorm replacement Co-authored-by: Philip Pilgerstorfer <[email protected]> * major changes * add standardization * minort changes * fix tests * rename reg parameters * linting * test coverage, docstting * check array for infs * fix isinstance to base type * fix isort, add test coverage * new tutorial * docstring fix Co-authored-by: Zain Patel <[email protected]> * test string match Co-authored-by: Zain Patel <[email protected]> * assert improvement Co-authored-by: Zain Patel <[email protected]> * SWE suggestions * minor bugfix * more test fixing Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * release.md, version bump, docs Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Jebq <[email protected]> * Release/0.8.1 (#94) * Merge back to develop * Simplifying viz.draw syntax in tutorial notebook (#46) * Add non negativity constraint in numpy lasso (#41) * Add plotting tutorial to the documentation (#47) * Unpin some requirements * Mixed type data generation (#55) Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Merge back to develop (#59) * Pytorch NOTEARS (#63) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Update release.md * Address coments * Corrected release.md * fc1 to adjacency Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> * Pinned sphinx-auto-doc-typehints (#66) * Corrected a spelling/grammar mistake (#55) * Fix/lint (#73) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * black * pin pytorch version * pin pytorch version Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Structure learning regressor (#68) * initial commit (local copy-paste) * fixed minor comments * minor bugfix * impute from children inital commit * bugfixes and method option * auto thresholding * autothreshold and bugfix * make threshold removal explicit * add l1 argument * remove child imputation * feat importance fix and tabu logic * moved threshold till dag * restructure with base class * coef mask * recipe * enable bias fitting * persist bias as node attribute * allow fit_intercept * minor PR comment fixes * minor comment adjustment * test coverage and l1 clarification * recipe * minor test fixes * more tests * full test coverage * revove python 3.5/3.6 unsupported import * add normalization option * idiomatic typing * correct pylint errors * update some tests * more typeing updates * more pylint requirements * more pylint disable * python 3.5 support * try to get to work with 3.5 * full coverage and 3.5 support * remove base class to pass test * remove unneeded supression * black formatting changes * remove unused import * pytlint supression * minor reformat change * isort fix * better defensive programming * fix unittests * docstring update * do Raises docstring properly * action SWE suggestions * hotfixes * minor update * minor black formatting change * final merge checkbox * fix end of file * Data Gen root node initialisation fix (#72) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * fix for consinuous normal data * generalise across all dtypes * support fit_intercept * fixed many test errors * test logic fixes * lint test fixes * python 3.5 failure change * minor test bugfix * black * pin pytorch version * pin pytorch version * additional test parameter * black formatting * requested changes * test updates and docstring * black format change * disable too many lines * change * move recipe to tutorial folder * releaseMD changes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: qbphilip <[email protected]> * [1/2] Poisson data for data gen (#61) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * variable name change bugfix * fix tests Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> * [2/2] Nonlinear Data gen (#60) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * refactor & docstring * remove unused helper object * add data gen to init * make test more robust * add count data and test, use logs for poisson samples for stability * add nonlinear * fix tests * duplicate fixtures * remove unused fixtures * refactor data_generators into package with core and wrappers * move wrapper to test_wrapper * add nonlinear to init * change order in all * change release.md * root node fix on core + count * nonlinear support to wrappers * docstring update * bugfix and reproducability fix * many tests and test updates * poiss bugfix and test fix * moar test coverage * categorical dataframe test coverage * full test coverage and linting * fix linting and fstring * black reformat * fix unused pylint argument * pytest fix * FINAL linting fix * Fix stuff (#75) CircleCI fixes Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Zain Patel <[email protected]> * update black version (#76) * fix black * Fix/check for NA or Infinity when notears is used (#54) * update scipy version (#77) * add DYNOTEARS implementation (#50) Adds DYNOTEARS and corresponding data generator (for testing) * Pytorch NOTEARS extension - Non-Linear/Hidden Layer (#65) * NoTears as ScoreSolver * refactor continuous solver * adding attribute to access weight matrix * refactoring continuous solver * Adding fit_lasso method * add data_gen_continuous.py and tests (#38) * add data_gen.py * rename * wrap SM * move data_gen_continous, create test * more coverage * test fixes * move discrete sem to another file * node list dupe check test * ValueError tests * replace dag and sem functions with Ben's verions * add Ben's tests * fix fstring * to_numpy_array coverage * Ben's comments * remove unreachable ValueError for coverage * remove unused fixture * remove redundant test * remove extensions Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docstring Co-Authored-By: Ben Horsburgh <[email protected]> * docs Co-Authored-By: Ben Horsburgh <[email protected]> * doc Co-Authored-By: Ben Horsburgh <[email protected]> * rename file, g_dag rename to sm * add new tests for equal weights * docstring * steve docstring, leq fix * steve comments + docstrings Co-authored-by: Ben Horsburgh <[email protected]> * Adding check input and removing some inner functions * Removing attribute original_ndarray * Aligning from pandas with new implementation * Adding tests for fit_lasso * More tests for lasso * wrapping tabu params in a dict * Aligning tests with new tabu params * Aligning from_pandas with new tabu_params * Adding fit_intercept option to _fit method * Adding scaling option * fixing lasso tests * Adding a test for fit_intercept * scaling option only with mean * Correction in lasso bounds * Fix typos * Remove duplicated bounds function * adding comments * add torch files from xunzheng * add from_numpy_torch function that works like from_numpy_lasso * lint * add requirements * add debug functionality * add visual debug test * add license * allow running as main for viz, comments * move to contrib * make multi layer work a bit better * add comment for multi layer * use polynomial dag constraint for better speed comparison * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * revert unnecessary changes to keep PR lean * fixes * refactor * Integrated tests * Checkpoint * Refactoring * Finished initial refactoring * All tests passed * Cleaning * Git add testing * Get adjacency matrix * Done cleaning * Revert change to original notears * Revert change to original structuremodel * Revert change to pylintrc * Undo deletion * Apply suggestions from Zain Co-authored-by: Zain Patel <[email protected]> * Addressed Zain comments * Migrated from_numpy * Delete contrib test * Migrated w_threshold * Some linting * Change to None * Undo deletion * List comprehension * Refactoring scipy and remove scipy optimiser * Refactoring * Refactoring * Refactoring complete * change from np to torch tensor * More refactoring * Remove hnew equal to None * Refactor again and remove commented line * Minor change * change to params * Addressing Philip's comment * Add property * Add fc2 property weights * Change to weights * Docstring * Linting * Linting completed * Add gpu code * Add gpu to from_numpy and from_pandas * cuda 0 run out of memory * Debugging * put 5 * debugging gpu * shift to inner loop * debugging not in place * Use cada instead of to * Support both interfaces * Benchmarking gpu * Minor fix * correct import path for test * change gpu from 5 to 1 * Debugging * Debugging * Experimenting * Linting * Remove hidden layer and gpu * Linting * Testing and linting * Correct pytorch to torch * Add init zeros * Change weight threshold to 0.25 * Revert requirements.txt * Add hidden layer * small refactor * directional adj * minor edits * fix bias issues * breaking changes update to the interface * typo * new regressor regularisation interface * update forward method * forward(X) predictions work * working! * bugfix data normalisation * some fixes * average regularisation and adj calc at end * give credit! Co-authored-by: Philip Pilgerstorfer <[email protected]> * loc lin docstring update Co-authored-by: Philip Pilgerstorfer <[email protected]> * docstring + fc1/fc2 name updates * moar docstring updates * more minor updates * remove normalize option * plotting util * rename to DAGRegressor * rename and checks * more util functions * fix bias * fix bias with no intercept * fix linear adj * add tests * minor fix * minor fixes * extend interface to bias * differentialte coef_ and feature_imporances * seperate bias element * tests * more test coverage * nonlinear test coverage * test hotfix * more test coverage * test requirements update * more test coverage * formatting changes * final pylint change * more linting * more bestpractice structuring * more minor fixes * FINAL linting updates * actual last change * update to reg defaults, additions to the tutorial * nonlinear regularisation updates * regressor tutorial * almost finishing touches * gradient based h function! * soft clamp and coef feature importance seperation * small api update, closer to batchnorm * docstring updates * stronger soft clamping * gradient L1 rather than L2 * fcpos neg removal, gradient optim * revert back to create_graph=True for 2nd derivative * remove print and test fix * black reformatting * new black version * full test coverage * isort fix * pylint fix * first layer h(W) for speed optimization * fix batch norm system * add nonlinear test * test hotfix * black reformat * isort fix * remove X requirement from h_func * regressor tutorial final commit and black update * LayerNorm replacement Co-authored-by: Philip Pilgerstorfer <[email protected]> * major changes * add standardization * minort changes * fix tests * rename reg parameters * linting * test coverage, docstting * check array for infs * fix isinstance to base type * fix isort, add test coverage * new tutorial * docstring fix Co-authored-by: Zain Patel <[email protected]> * test string match Co-authored-by: Zain Patel <[email protected]> * assert improvement Co-authored-by: Zain Patel <[email protected]> * SWE suggestions * minor bugfix * more test fixing Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: angeldrothqb <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * Merge release/0.8.0 back into develop for release 0.8.0 (#82) * Hotfix/0.4.3 (#7) - Address broken links and grammar * Fix documentation links in README (#2) * Fix links in README * library -> libraries * Fix github link in docs * Clean up grammar and consistency in documentation (#4) * Clean up grammar and consistency in `README` files * Add esses, mostly * Reword feature description to not appear automatic * Update docs/source/05_resources/05_faq.md Co-Authored-By: Ben Horsburgh <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> * hotfix/0.4.3: fix broken links Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> * Release/0.5.0 * Plotting now backed by pygraphviz. This allows: * More powerful layout manager * Cleaner fully customisable theme * Out-the-box styling for different node and edge types * Can now get subgraphs from StructureModel containing a specific node * Bugfix to resolve issue when fitting CPDs with some missing states in data * Minor documentation fixes and improvements * Release/0.6.0 * Release/0.7.0 (#57) * Added plottting tutorial to the documentation * Updated `viz.draw` syntax in tutorial notebooks * Bugfix on notears lasso (`from_numpy_lasso` and `from_pandas_lasso`) where the non-negativity constraint was not being set * Added DAG-based synthetic data generator for mixed types (binary, categorical, continuous) using a linear SEM approach. * Unpinned some requirements * release.md, version bump, docs Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> * fix tests (#87) * Fix pygments fail (#84) Thanks Zain! * update notebook beginning (#89) * Add Binary distribution type support (#85) * binary dtype folder and __init__ * dtype base class * continuous dtype * binary dtype * update core * make plural * update interface for idx * minor variable name change * notears update * python 3.5 support * fix fstring * remove categorical methods, doctrings * formatting and docstrings * remove redundant cat code * isort * indexerror check * defensive check tests * datatype loss tests * more test coverage * more tests and formatting * fix test import * remove double test * linting * docstring and pylint * docstring fix Co-authored-by: Zain Patel <[email protected]> * fix long string Co-authored-by: Zain Patel <[email protected]> * docstring fix Co-authored-by: Zain Patel <[email protected]> * remove relative imports * docstring fix * dict comprehension * list comprehension and neatness * remove unuesd import to __init__ * fix test * remove unused return interface * add binary f1score tests * one datatype instane per feature * rename dtype -> disttype, attach dists to nodes * fix tests * fix linting * fix preserve node dtyper * fix tests * fix tests * fix tests * final docstring and test fixes * lint fix * test_fix, warning * linting * fix test * fix tests * reduce threshold of test * raise better error * black linting * remove warning * remove useless supression and import Co-authored-by: Zain Patel <[email protected]> * Add sklearn binary classifier (#90) * binary dtype folder and __init__ * dtype base class * continuous dtype * binary dtype * update core * make plural * update interface for idx * minor variable name change * notears update * python 3.5 support * fix fstring * remove categorical methods, doctrings * formatting and docstrings * remove redundant cat code * isort * indexerror check * defensive check tests * datatype loss tests * more test coverage * more tests and formatting * fix test import * remove double test * linting * docstring and pylint * docstring fix Co-authored-by: Zain Patel <[email protected]> * fix long string Co-authored-by: Zain Patel <[email protected]> * docstring fix Co-authored-by: Zain Patel <[email protected]> * remove relative imports * docstring fix * dict comprehension * list comprehension and neatness * remove unuesd import to __init__ * fix test * remove unused return interface * new sklearn folder structure * sklearn class outline * new dtype interface * docstring clarification * inverse link function * add binary f1score tests * one datatype instane per feature * rename dtype -> disttype, attach dists to nodes * fix tests * fix linting * fix preserve node dtyper * fix tests * fix tests * fix tests * final docstring and test fixes * lint fix * test_fix, warning * linting * fix test * fix tests * reduce threshold of test * docstring clarification * _target_dist_type injection * docstring updates + clf fit outline * old doctring deprecation * clf predict_proba and predict * return bugfix * docstring update * import fix, linting, clf fit finished * args docstring and None schema * raise better error * black linting * linting * revert to public interface * remove warning * remove useless supression and import * add useless change to resolve merge conflict * update inits * standardization and data reconstruction * remove unused imports * fix clf .precit() * regressor fit_predict * remove useless regressor predict * test import fix * fix warnings * pass series name thru * fix schema pass thru * better dict comprehension Co-authored-by: Zain Patel <[email protected]> * import and comment fixes * update to .format() * fig sklearn is fitted test * more dtype schema insertion * DAGRegressor test fix * dag regressor test * more linting * big test restructure * combined test suite * error string update * more test coverage * linting and isort * move test to combined test * return float64 preds * moar clf tests * remove untestable (multiclass) code * class number error test * balck reformat * docstrings, pylint * fix test bug * standard scaler for _base * pull classes direct from LabelEncoder * update tutorial Co-authored-by: Zain Patel <[email protected]> * Fix/uniform discretisation (#65) * Fix uniform discretiser * Fix uniform discretiser Co-authored-by: Philip Pilgerstorfer <[email protected]> * release.MD * version bump * Update causalnex/structure/pytorch/dist_type/_base.py Co-authored-by: Zain Patel <[email protected]> * Update causalnex/structure/pytorch/dist_type/__init__.py Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: Philip Pilgerstorfer <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: qbphilip <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Jebq <[email protected]> Co-authored-by: Shuhei Ishida <[email protected]> Co-authored-by: Ben Horsburgh <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: Nikos Tsaousis <[email protected]> Co-authored-by: Deepyaman Datta <[email protected]> Co-authored-by: Angel Droth <[email protected]> Co-authored-by: GabrielAzevedoFerreiraQB <[email protected]> Co-authored-by: stevelersl <[email protected]> Co-authored-by: LiseDiagneQB <[email protected]> Co-authored-by: Casey Juanxi Li <[email protected]> Co-authored-by: Zain Patel <[email protected]> Co-authored-by: KING-SID <[email protected]> Co-authored-by: Jebq <[email protected]> Co-authored-by: Shuhei Ishida <[email protected]>

ziyuwzf closed this as completed Jul 30, 2020

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

I want to find some causal relationship from some cols to label what i defined by from_pandas(no tears),but i got some causal relationship from label to others #63

I want to find some causal relationship from some cols to label what i defined by from_pandas(no tears),but i got some causal relationship from label to others #63

ziyuwzf commented Jul 30, 2020

I want to find some causal relationship from some cols to label what i defined by from_pandas(no tears),but i got some causal relationship from label to others #63

I want to find some causal relationship from some cols to label what i defined by from_pandas(no tears),but i got some causal relationship from label to others #63

Comments

ziyuwzf commented Jul 30, 2020

Description

Context