Feature/mds example #3728
Feature/mds example #3728
Conversation
|
@ElenaNST ok those errors on travis are actually real errors with this patch. |
| @@ -105,3 +105,13 @@ Neural Networks | |||
| :caption: Neural Netwroks | |||
|
|
|||
| examples/neural_nets/** | |||
|
|
|||
| Feature Transformations | |||
There was a problem hiding this comment.
Maybe this should be "Feature converters"
There was a problem hiding this comment.
Glad to change it, but are you sure about it? I haven't come across the term "Feature converters" in the machine learning literature.
There was a problem hiding this comment.
I guess we call these things "converter" internally.
Maybe just leave it as is now, easy to change later
doc/cookbook/source/references.bib
Outdated
| @@ -153,3 +153,15 @@ @article{Quinonero-Candela2005 | |||
| year = {2005}, | |||
| pages = {1939--1959}, | |||
| } | |||
| @TECHREPORT{Wickelmaier03anintroduction, | |||
| author = {Florian Wickelmaier}, | |||
doc/cookbook/source/references.bib
Outdated
| @@ -153,3 +153,15 @@ @article{Quinonero-Candela2005 | |||
| year = {2005}, | |||
| pages = {1939--1959}, | |||
| } | |||
| @TECHREPORT{Wickelmaier03anintroduction, | |||
| author = {Florian Wickelmaier}, | |||
| title = {An introduction to MDS}, | |||
There was a problem hiding this comment.
We need a better reference here. Maybe David Barbers book?
There was a problem hiding this comment.
If you are referring to Bayesian Reasoning and Machine learning, there's no reference to MDS. Maybe Modern Multidimensional Scaling from Borg is ok? (also cited in the documentation of CMultidimensionalScaling)
doc/cookbook/source/references.bib
Outdated
| } | ||
| @inproceedings{Silva2004SparseMS, | ||
| title={Sparse multidimensional scaling using landmark points}, | ||
| author={Vin de Silva and Joshua B. Tenenbaum}, |
doc/cookbook/source/references.bib
Outdated
| @inproceedings{Silva2004SparseMS, | ||
| title={Sparse multidimensional scaling using landmark points}, | ||
| author={Vin de Silva and Joshua B. Tenenbaum}, | ||
| booktitle={}, |
There was a problem hiding this comment.
watch you bibtex style, this needs a conference name (check the others to see how we name them)
| .. math:: | ||
| Strain_D(x_1, x_2, \cdots, x_N ) = \frac{\sum_{ij} (b_{ij} -d_{ij})^2}{\sum_{ij} b_{ij}^2} | ||
|
|
||
| where :math:`b_{ij}` are the terms of matrix :math:`B`, derived from matrix :math:`D` by applying double centering. This problem is tractable and solved using eigenvalue decomposition from :math:`B=X X'` |
There was a problem hiding this comment.
Transpose is ^\top in latex.
Please double check the math and make sure every quantity is defined, and used somewhere.
|
|
||
| where :math:`b_{ij}` are the terms of matrix :math:`B`, derived from matrix :math:`D` by applying double centering. This problem is tractable and solved using eigenvalue decomposition from :math:`B=X X'` | ||
|
|
||
| :cite:`Silva2004SparseMS` introduced a two-step, computationally efficient approximation of classical MDS for use when the number of data points is too large. First, it uses a subset of the initial points as landmarks to apply classical MDS and then, it uses a distance-based triangulation procedure to determine the positions of the remaining points. |
There was a problem hiding this comment.
Is this demonstrated in here? If so, say so, if not it can be removed
There was a problem hiding this comment.
It is not demonstrated, but this converter.set_landmark(False) deactivates it. So, if I remove this paragraph I should write something like "We deactivate an optimization technique". I'm afraid that's a bit vague and I didn't find any relevant case in the cookbooks, so what's your approach? How about just dropping the second sentence?
There was a problem hiding this comment.
I would explicitly demostrate this in a second step in the example. First without and then with, mentioning it
|
|
||
| .. sgexample:: multidimensional_scaling:create_features | ||
|
|
||
| We calculate the Euclidean distance between our features. |
There was a problem hiding this comment.
use sgclass for Euclidean distance here
|
|
||
| .. sgexample:: multidimensional_scaling:calculate_distance_matrices | ||
|
|
||
|
|
| #![calculate_distances_after] | ||
|
|
||
| #![calculate_distance_matrices] | ||
| RealMatrix distance_matrix_after = distance_after.get_distance_matrix() |
There was a problem hiding this comment.
move these to calculate_distances_before and calculate_distances_after
No need for its own section
eleninisioti
force-pushed
the
feature/MdsExample
branch
from
2f34728 to
205faf8
Compare
|
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. |
|
This issue is now being closed due to a lack of activity. Feel free to reopen it. |
I have replaced the "converter_multidimensionalscaling_modular.py" example with a meta example and a cookbook page. I provide the .sg and .rst files.
Attention: I am not sure about line 30 of the .sg file. First of all, I am not sure if it should be included and even if it should it does not compile, as make complains about the
\and>symbols.I've also commented here as I'm not sure where the cookbook should be placed, but I think this is a better place to discuss about it.