Derandomized + Adaptive Classical Shadows #111

alexfleury-sb · 2022-01-11T21:54:41Z

Derandomized + adaptive flavors of classical shadows.

I moved everything to a separate folder to keep the toolboxes.measurements cleaner.
There are some places that could be done differently, but I took some editorial choices. Those could be changed (I'm thinking of a map(lambda..) somewhere.
I removed the estimate_state from the requirements. I don't think it is necessary to implement this in every shadow as it defeats the purpose of constructing the shadow. But this is something that can be added in the future.
Some functions are identical in DerandomizedClassicalShadow and AdaptiveClassicalShadow. Using inheritance can prevent this but at the cost of making the code harder to follow in my opinion.

tangelo/toolboxes/measurements/classical_shadows/classical_shadows.py

ValentinS4t1qbit

I think these comments may help you grow quite a bit as a developer, as you worked with quite complex logic but got tangled up in it, and ended up with something confusing and convoluted.

Some methods were in fact so complicated that I will need to review it again later when they're a bit cleaner and may need an explanation as to what they intend to do :)

tangelo/toolboxes/measurements/__init__.py

tangelo/toolboxes/measurements/classical_shadows/adaptive.py

tangelo/toolboxes/measurements/classical_shadows/derandomized.py

ValentinS4t1qbit

A little fix

ValentinS4t1qbit

Some more work :)

A few good things to learn, I'd say (list and dict, copy vs deepcopy...)

tangelo/toolboxes/measurements/classical_shadows/adaptive.py

tangelo/toolboxes/measurements/classical_shadows/derandomized.py

ValentinS4t1qbit

Good job !

* Sphinx docs (#76) * Fixing CCSDSolver.get_rdm() with frozen orbitals and energy from RDMs (#81) * small fixes to allow initial density matrix for faster noisy sampling with cirq (#84) * Branding (Tangelo, Good Chemistry Company) (#87) * Name change: backendbuddy -> linq (#93) * added multi-controls multi-targets, extra gates (#88) * Add QMF and QCC capabilities and tests (#91) * JKMN mapping implementation (#95) * added inverse function to Circuit (#78) * added pycodestyle tests (#96) * Circuit methods (repetition operator, equality, trim and split methods) (#101) * Staged controlled time (#100) * Support for name attribute in Circuit class (#110) * Derandomized + Adaptive Classical Shadows (#111) * added vsqs ansatz (#109) * Staged richardson (#99) * Majorana pool for ADAPT (#114) * Copy gate data instead of referencing it when instantiation Circuit object (#118) * Fixed QEMIST Cloud QPU connection ctrl-c in job_result. (#121) * Estimate QPU cost with QEMIST Cloud API. (#120) * Improvements for handling exp data with ClassicalShadow (#124) * Qulacs operator build changed to fix memory leak (#122) Co-authored-by: ValentinS4t1qbit <[email protected]> Co-authored-by: AlexandreF-1qbit <[email protected]> Co-authored-by: James Brown <[email protected]> Co-authored-by: JamesB-1qbit <[email protected]> Co-authored-by: MPCoons <[email protected]> Co-authored-by: elloyd-1qbit <[email protected]> Co-authored-by: KrzysztofB-1qbit <[email protected]> Co-authored-by: Rudi Plesch <[email protected]> Co-authored-by: GitHub Actions <[email protected]>

* v0.3.0 into main (#127) * Sphinx docs (#76) * Fixing CCSDSolver.get_rdm() with frozen orbitals and energy from RDMs (#81) * small fixes to allow initial density matrix for faster noisy sampling with cirq (#84) * Branding (Tangelo, Good Chemistry Company) (#87) * Name change: backendbuddy -> linq (#93) * added multi-controls multi-targets, extra gates (#88) * Add QMF and QCC capabilities and tests (#91) * JKMN mapping implementation (#95) * added inverse function to Circuit (#78) * added pycodestyle tests (#96) * Circuit methods (repetition operator, equality, trim and split methods) (#101) * Staged controlled time (#100) * Support for name attribute in Circuit class (#110) * Derandomized + Adaptive Classical Shadows (#111) * added vsqs ansatz (#109) * Staged richardson (#99) * Majorana pool for ADAPT (#114) * Copy gate data instead of referencing it when instantiation Circuit object (#118) * Fixed QEMIST Cloud QPU connection ctrl-c in job_result. (#121) * Estimate QPU cost with QEMIST Cloud API. (#120) * Improvements for handling exp data with ClassicalShadow (#124) * Qulacs operator build changed to fix memory leak (#122) Co-authored-by: ValentinS4t1qbit <[email protected]> Co-authored-by: AlexandreF-1qbit <[email protected]> Co-authored-by: James Brown <[email protected]> Co-authored-by: JamesB-1qbit <[email protected]> Co-authored-by: MPCoons <[email protected]> Co-authored-by: elloyd-1qbit <[email protected]> Co-authored-by: KrzysztofB-1qbit <[email protected]> Co-authored-by: Rudi Plesch <[email protected]> Co-authored-by: GitHub Actions <[email protected]>

* Bump develop to v0.3.0 (#127) (#128) * v0.3.0 into main (#127) * Sphinx docs (#76) * Fixing CCSDSolver.get_rdm() with frozen orbitals and energy from RDMs (#81) * small fixes to allow initial density matrix for faster noisy sampling with cirq (#84) * Branding (Tangelo, Good Chemistry Company) (#87) * Name change: backendbuddy -> linq (#93) * added multi-controls multi-targets, extra gates (#88) * Add QMF and QCC capabilities and tests (#91) * JKMN mapping implementation (#95) * added inverse function to Circuit (#78) * added pycodestyle tests (#96) * Circuit methods (repetition operator, equality, trim and split methods) (#101) * Staged controlled time (#100) * Support for name attribute in Circuit class (#110) * Derandomized + Adaptive Classical Shadows (#111) * added vsqs ansatz (#109) * Staged richardson (#99) * Majorana pool for ADAPT (#114) * Copy gate data instead of referencing it when instantiation Circuit object (#118) * Fixed QEMIST Cloud QPU connection ctrl-c in job_result. (#121) * Estimate QPU cost with QEMIST Cloud API. (#120) * Improvements for handling exp data with ClassicalShadow (#124) * Qulacs operator build changed to fix memory leak (#122) Co-authored-by: ValentinS4t1qbit <[email protected]> Co-authored-by: AlexandreF-1qbit <[email protected]> Co-authored-by: James Brown <[email protected]> Co-authored-by: JamesB-1qbit <[email protected]> Co-authored-by: MPCoons <[email protected]> Co-authored-by: elloyd-1qbit <[email protected]> Co-authored-by: KrzysztofB-1qbit <[email protected]> Co-authored-by: Rudi Plesch <[email protected]> Co-authored-by: GitHub Actions <[email protected]> * Recomputing MF when working with atom indices in DMET (#130) * Recomputing and testing MF when working with atom indexes. Change Localization import level. * More gates for openqasm translator (#129) * Added CZ, CY, CRz, CP, CSWAP and SWAP to openqasm translator * fixed return for return_phase=False in trotterize (#133) * fixed return for return_phase=False in trotterize * ONIOM problem reformulation (#119) * Updated ONIOM notebook use case * QubitHamiltonian get_operators bug (#131) * Fixed get_operators bug + added a test. * Measurement map (#134) * Measurement map + extending qwc partioning with repetition * Working state, cleaning is wip. * Givens gate (#135) * added givens gate decomposition Co-authored-by: ValentinS4t1qbit <[email protected]> * Notebook classical shadow (#123) Classical shadows notebook. It has been added to tests as well Co-authored-by: ValentinS4t1qbit <[email protected]> * added class to prepare or decompute an arbitrary statevector (#137) * Many small todos (#138) * Add an explanation how Circuit.reindex_qubits method is working. * ClassicalShadows.simulate only unique circuit (massive speedup). * Better management of backend options in VQESolver. * CS notebook update. * Comment fix for new return. * No need of n_electrons with ref_state==zero. * Change method for simulate (noisy?). * Diag coulomb (#136) * diagonalizing circuits implemented and tested * improvements to jkmn leaf->majorana selection (#139) * improvements to jkmn leaf->majorana selection * Use of get_vector function. * Change constructors -> classmethod. * Change interface + working state. * Stage where I have to write tests. * Test for HybridOperator and Z2 tapering. * Conformance tests + typos. * Moved file + typos. * Added tests for matrix manip. * Full path of load_operator in tests. * Permit all active orbitals partially occupied (#146) * ONIOM capping with chemical groups (#141) * Implementation for other chemical groups capping in ONIOM. Tests, docs. * uccgd ansatz (#144) * uccgd ansatz for use in SA-OO-VQE * First round + added spin=\=0 fix. * Support for symmetry, now call pyscf directly (#147) * call pyscf directly, symmetry now supported * Rotosolve implementation as an optimizer for parameterized circuits (#142) * rotosolve implementation and tests * ONIOM multisolvers (#143) * Multisolver support in ONIOM * Added others solvers to ONIOM, added tests and ROHF support (RHF->HF). * sa_vqe_solver from sa-oo-vqe branch * added import to __init__ * fixes for PR * clearer documentation * change statevector jkmn * replaced jkmn_prep_circuit with jkmn_prep_vector * Fixed error message. * Some reviews. * JKMN z2 tapering tests. * Docstrings in tests. * Speed improvement + bitwise operations + moved collapse function. * Passing conformance tests. * added state averaged orbital optimized files * merged molecule symmetry changes * Update test_hybridoperator.py aligning. * Hybrid -> Multiform. * Small fixes. * PR changes * support for reading in xyz files (#151) * support for reading in xyz files * Update z2_tapering.py * Update sa_oo_vqe_solver.py * Ilc ansatz (#132) * Add ILC ansatz class * updates to qmf, qcc, and ilc ansatz classes * enable QMF and ILC classes to read-in and process data from OpenFermion *.hdf5 files and other small fixes. Co-authored-by: ValentinS4t1qbit <[email protected]> * Updating information, and adding Windows install comment (#155) * Quantum deflation (#152) * added deflation * added ref_states as circuit and to VQESolver * added ref_state and deflation to adapt vqe Co-authored-by: ValentinS4t1qbit <[email protected]> * changed default basis to crenbl in pyscf, added ecp option (#156) * changed default basis to crenbl for pyscf to retrieve number of electrons & atoms, added ecp option * Cancel redundant gates (#153) * Functions / methods to remove small rotations, and remove gates that cancel each other * Circuit depth (#159) * Depth method for Circuit class * MI-FNO link (#157) * MI-FNO helper class. Added tests + reconstruction of MI energy, support for building sermonic operators for fragments * Iqcc solver (#154) * iQCC solver, frobenius norm compression method on QubitOperator Co-authored-by: ValentinS4t1qbit <[email protected]> * MIFNO notebook (#161) * MIFNO notebook. Added to tests and sphinx docs Co-authored-by: Valentin Senicourt <[email protected]> Co-authored-by: ValentinS4t1qbit <[email protected]> * Docs fix + tutorials removal from sphinx docs (#162) * Docs: requirements.txt no longer needed. Tutorials removed from generated docs (redundant with Github, issues with latex, looks better on github and will display images once repo is public Co-authored-by: AlexandreF-1qbit <[email protected]> * readme upgrade (#164) * readme upgrade and images, contributions.rst for code of conduct mention * Fixes for docs and readme, files for Pypi (#166) * Docs and readme fixed with feedback. Pypi file removed as it hinders the installation from source process: a guide will be provided to maintainer team for pypi update. Co-authored-by: AlexandreF-1qbit <[email protected]> * Version number and changelogs bumped Co-authored-by: AlexandreF-1qbit <[email protected]> Co-authored-by: James Brown <[email protected]> Co-authored-by: JamesB-1qbit <[email protected]> Co-authored-by: MPCoons <[email protected]> Co-authored-by: elloyd-1qbit <[email protected]> Co-authored-by: KrzysztofB-1qbit <[email protected]> Co-authored-by: Rudi Plesch <[email protected]> Co-authored-by: GitHub Actions <[email protected]>

* v0.3.0 into main (#127) * Sphinx docs (#76) * Fixing CCSDSolver.get_rdm() with frozen orbitals and energy from RDMs (#81) * small fixes to allow initial density matrix for faster noisy sampling with cirq (#84) * Branding (Tangelo, Good Chemistry Company) (#87) * Name change: backendbuddy -> linq (#93) * added multi-controls multi-targets, extra gates (#88) * Add QMF and QCC capabilities and tests (#91) * JKMN mapping implementation (#95) * added inverse function to Circuit (#78) * added pycodestyle tests (#96) * Circuit methods (repetition operator, equality, trim and split methods) (#101) * Staged controlled time (#100) * Support for name attribute in Circuit class (#110) * Derandomized + Adaptive Classical Shadows (#111) * added vsqs ansatz (#109) * Staged richardson (#99) * Majorana pool for ADAPT (#114) * Copy gate data instead of referencing it when instantiation Circuit object (#118) * Fixed QEMIST Cloud QPU connection ctrl-c in job_result. (#121) * Estimate QPU cost with QEMIST Cloud API. (#120) * Improvements for handling exp data with ClassicalShadow (#124) * Qulacs operator build changed to fix memory leak (#122) Co-authored-by: ValentinS4t1qbit <[email protected]> Co-authored-by: AlexandreF-1qbit <[email protected]> Co-authored-by: James Brown <[email protected]> Co-authored-by: JamesB-1qbit <[email protected]> Co-authored-by: MPCoons <[email protected]> Co-authored-by: elloyd-1qbit <[email protected]> Co-authored-by: KrzysztofB-1qbit <[email protected]> Co-authored-by: Rudi Plesch <[email protected]> Co-authored-by: GitHub Actions <[email protected]> * Update _version.py * Update to release/v0.3.1 (#167) * Bump develop to v0.3.0 (#127) (#128) * v0.3.0 into main (#127) * Sphinx docs (#76) * Fixing CCSDSolver.get_rdm() with frozen orbitals and energy from RDMs (#81) * small fixes to allow initial density matrix for faster noisy sampling with cirq (#84) * Branding (Tangelo, Good Chemistry Company) (#87) * Name change: backendbuddy -> linq (#93) * added multi-controls multi-targets, extra gates (#88) * Add QMF and QCC capabilities and tests (#91) * JKMN mapping implementation (#95) * added inverse function to Circuit (#78) * added pycodestyle tests (#96) * Circuit methods (repetition operator, equality, trim and split methods) (#101) * Staged controlled time (#100) * Support for name attribute in Circuit class (#110) * Derandomized + Adaptive Classical Shadows (#111) * added vsqs ansatz (#109) * Staged richardson (#99) * Majorana pool for ADAPT (#114) * Copy gate data instead of referencing it when instantiation Circuit object (#118) * Fixed QEMIST Cloud QPU connection ctrl-c in job_result. (#121) * Estimate QPU cost with QEMIST Cloud API. (#120) * Improvements for handling exp data with ClassicalShadow (#124) * Qulacs operator build changed to fix memory leak (#122) Co-authored-by: ValentinS4t1qbit <[email protected]> Co-authored-by: AlexandreF-1qbit <[email protected]> Co-authored-by: James Brown <[email protected]> Co-authored-by: JamesB-1qbit <[email protected]> Co-authored-by: MPCoons <[email protected]> Co-authored-by: elloyd-1qbit <[email protected]> Co-authored-by: KrzysztofB-1qbit <[email protected]> Co-authored-by: Rudi Plesch <[email protected]> Co-authored-by: GitHub Actions <[email protected]> * Recomputing MF when working with atom indices in DMET (#130) * Recomputing and testing MF when working with atom indexes. Change Localization import level. * More gates for openqasm translator (#129) * Added CZ, CY, CRz, CP, CSWAP and SWAP to openqasm translator * fixed return for return_phase=False in trotterize (#133) * fixed return for return_phase=False in trotterize * ONIOM problem reformulation (#119) * Updated ONIOM notebook use case * QubitHamiltonian get_operators bug (#131) * Fixed get_operators bug + added a test. * Measurement map (#134) * Measurement map + extending qwc partioning with repetition * Working state, cleaning is wip. * Givens gate (#135) * added givens gate decomposition Co-authored-by: ValentinS4t1qbit <[email protected]> * Notebook classical shadow (#123) Classical shadows notebook. It has been added to tests as well Co-authored-by: ValentinS4t1qbit <[email protected]> * added class to prepare or decompute an arbitrary statevector (#137) * Many small todos (#138) * Add an explanation how Circuit.reindex_qubits method is working. * ClassicalShadows.simulate only unique circuit (massive speedup). * Better management of backend options in VQESolver. * CS notebook update. * Comment fix for new return. * No need of n_electrons with ref_state==zero. * Change method for simulate (noisy?). * Diag coulomb (#136) * diagonalizing circuits implemented and tested * improvements to jkmn leaf->majorana selection (#139) * improvements to jkmn leaf->majorana selection * Use of get_vector function. * Change constructors -> classmethod. * Change interface + working state. * Stage where I have to write tests. * Test for HybridOperator and Z2 tapering. * Conformance tests + typos. * Moved file + typos. * Added tests for matrix manip. * Full path of load_operator in tests. * Permit all active orbitals partially occupied (#146) * ONIOM capping with chemical groups (#141) * Implementation for other chemical groups capping in ONIOM. Tests, docs. * uccgd ansatz (#144) * uccgd ansatz for use in SA-OO-VQE * First round + added spin=\=0 fix. * Support for symmetry, now call pyscf directly (#147) * call pyscf directly, symmetry now supported * Rotosolve implementation as an optimizer for parameterized circuits (#142) * rotosolve implementation and tests * ONIOM multisolvers (#143) * Multisolver support in ONIOM * Added others solvers to ONIOM, added tests and ROHF support (RHF->HF). * sa_vqe_solver from sa-oo-vqe branch * added import to __init__ * fixes for PR * clearer documentation * change statevector jkmn * replaced jkmn_prep_circuit with jkmn_prep_vector * Fixed error message. * Some reviews. * JKMN z2 tapering tests. * Docstrings in tests. * Speed improvement + bitwise operations + moved collapse function. * Passing conformance tests. * added state averaged orbital optimized files * merged molecule symmetry changes * Update test_hybridoperator.py aligning. * Hybrid -> Multiform. * Small fixes. * PR changes * support for reading in xyz files (#151) * support for reading in xyz files * Update z2_tapering.py * Update sa_oo_vqe_solver.py * Ilc ansatz (#132) * Add ILC ansatz class * updates to qmf, qcc, and ilc ansatz classes * enable QMF and ILC classes to read-in and process data from OpenFermion *.hdf5 files and other small fixes. Co-authored-by: ValentinS4t1qbit <[email protected]> * Updating information, and adding Windows install comment (#155) * Quantum deflation (#152) * added deflation * added ref_states as circuit and to VQESolver * added ref_state and deflation to adapt vqe Co-authored-by: ValentinS4t1qbit <[email protected]> * changed default basis to crenbl in pyscf, added ecp option (#156) * changed default basis to crenbl for pyscf to retrieve number of electrons & atoms, added ecp option * Cancel redundant gates (#153) * Functions / methods to remove small rotations, and remove gates that cancel each other * Circuit depth (#159) * Depth method for Circuit class * MI-FNO link (#157) * MI-FNO helper class. Added tests + reconstruction of MI energy, support for building sermonic operators for fragments * Iqcc solver (#154) * iQCC solver, frobenius norm compression method on QubitOperator Co-authored-by: ValentinS4t1qbit <[email protected]> * MIFNO notebook (#161) * MIFNO notebook. Added to tests and sphinx docs Co-authored-by: Valentin Senicourt <[email protected]> Co-authored-by: ValentinS4t1qbit <[email protected]> * Docs fix + tutorials removal from sphinx docs (#162) * Docs: requirements.txt no longer needed. Tutorials removed from generated docs (redundant with Github, issues with latex, looks better on github and will display images once repo is public Co-authored-by: AlexandreF-1qbit <[email protected]> * readme upgrade (#164) * readme upgrade and images, contributions.rst for code of conduct mention * Fixes for docs and readme, files for Pypi (#166) * Docs and readme fixed with feedback. Pypi file removed as it hinders the installation from source process: a guide will be provided to maintainer team for pypi update. Co-authored-by: AlexandreF-1qbit <[email protected]> * Version number and changelogs bumped Co-authored-by: AlexandreF-1qbit <[email protected]> Co-authored-by: James Brown <[email protected]> Co-authored-by: JamesB-1qbit <[email protected]> Co-authored-by: MPCoons <[email protected]> Co-authored-by: elloyd-1qbit <[email protected]> Co-authored-by: KrzysztofB-1qbit <[email protected]> Co-authored-by: Rudi Plesch <[email protected]> Co-authored-by: GitHub Actions <[email protected]> Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: AlexandreF-1qbit <[email protected]> Co-authored-by: James Brown <[email protected]> Co-authored-by: JamesB-1qbit <[email protected]> Co-authored-by: MPCoons <[email protected]> Co-authored-by: elloyd-1qbit <[email protected]> Co-authored-by: KrzysztofB-1qbit <[email protected]> Co-authored-by: Rudi Plesch <[email protected]> Co-authored-by: GitHub Actions <[email protected]>

alexfleury-sb added 9 commits January 6, 2022 13:51

General cleaning, random imports? and reorg.

eb51fc5

Working state of derand + cleaning.

0b0a8e6

First cleanup of derandomized CS.

fd6814d

Monday morning commit.

4701ce0

Finish-ish derandomized + tests.

5e8b591

Removed obligation to define estimate_state.

0c5e7d7

Merge branch 'staging-0.3.0' into derandomized_cs.

aa57a8d

Added adaptive CS + cleaning.

01a4a01

General cleaning and docs.

1bd9d91

alexfleury-sb requested review from JamesB-1qbit and ValentinS4t1qbit as code owners January 11, 2022 21:54

Typo.

e4951f7

JamesB-1qbit reviewed Jan 12, 2022

View reviewed changes

tangelo/toolboxes/measurements/classical_shadows/classical_shadows.py Show resolved Hide resolved

ValentinS4t1qbit requested changes Jan 12, 2022

View reviewed changes

ValentinS4t1qbit reviewed Jan 12, 2022

View reviewed changes

alexfleury-sb and others added 12 commits January 12, 2022 17:11

Some fixes from the review.

83554ae

Removed unused code.

a95111b

Remove abs.

085594f

Adaptive ongoing work.

78ddea6

Changed the adaptive algo to make it consistent with the paper.

451f08a

Comments and updated tests.

3df9bcd

Merge branch 'staging-0.3.0' into derandomized_cs.

88b4d48

Removed auto import.

7ef81e9

Pycodestyle fixes.

fff0b52

General cleaning.

380b736

Comments cleaning.

a18e040

Update __init__.py

efc53e3

ValentinS4t1qbit requested changes Jan 14, 2022

View reviewed changes

2nd review pass.

56d822e

ValentinS4t1qbit approved these changes Jan 17, 2022

View reviewed changes

ValentinS4t1qbit merged commit 1b1f704 into staging-0.3.0 Jan 17, 2022

alexfleury-sb deleted the derandomized_cs branch January 18, 2022 16:09

ValentinS4t1qbit mentioned this pull request Feb 15, 2022

Bump develop to v0.3.0 (#127) #128

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Derandomized + Adaptive Classical Shadows #111

Derandomized + Adaptive Classical Shadows #111

alexfleury-sb commented Jan 11, 2022

ValentinS4t1qbit left a comment

ValentinS4t1qbit left a comment

ValentinS4t1qbit left a comment

ValentinS4t1qbit left a comment

Derandomized + Adaptive Classical Shadows #111

Derandomized + Adaptive Classical Shadows #111

Conversation

alexfleury-sb commented Jan 11, 2022

ValentinS4t1qbit left a comment

Choose a reason for hiding this comment

ValentinS4t1qbit left a comment

Choose a reason for hiding this comment

ValentinS4t1qbit left a comment

Choose a reason for hiding this comment

ValentinS4t1qbit left a comment

Choose a reason for hiding this comment