Add Gradient descent algorithms by paquiteau · Pull Request #196 · CEA-COSMIC/ModOpt

paquiteau · 2021-12-01T16:37:40Z

As announced in #194 , this PR adds support for gradient descent algorithms with various gradient direction and speed preconditioning.

* Add support for tensorflow backend which allows for differentiability (CEA-COSMIC#112) * Added support for tensorflow * Updates to get tests passing * Or --> And * Moving modopt to allow working with tensorflow * Fix issues with wos * Fix all flakes finally! * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Minute updates to codes * Add dynamic module * Fix docu * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Fix 115 (CEA-COSMIC#116) * Fix issues * Add right tests * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Minor bug fix, remove elif (CEA-COSMIC#124) Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Add tests for modopt.base.backend and fix minute bug uncovered (CEA-COSMIC#126) * Minor bug fix, remove elif * Add tests for backend * Fix tests * Add tests * Remove cupy * PEP fixes * Fix PEP * Fix PEP and update * Final PEP * Update setup.cfg Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update test_base.py Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Release cleanup (CEA-COSMIC#128) * updated GPU dependencies * added logo to manifest * updated package version and release date Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com>

sfarrens · 2021-12-02T10:45:35Z

Hi @paquiteau thanks for the suggestion and the PR! I will review this properly once all the CI issues are resolved. If you are having any problem with some of the WPS errors let me know as we can relax some errors for certain files. In particular, it will be necessary for algorithms/__init_.py in order to support the * imports.

paquiteau · 2021-12-02T14:51:10Z

I fixed most of the issues with linting, and updated setup.cfg when I could not find another way. I still have to write some test specific to the gradient descent though

modopt/opt/algorithms/gradient_descent.py

modopt/tests/test_opt.py

sfarrens · 2021-12-08T15:25:24Z

Hi @paquiteau!

I made a first pass on your PR. Firstly, very nice and meticulous work 👏 .

I proposed some minor changes to gradient_descent.py just to clean things up a bit, but otherwise the implementation looks good and the tests pass. The main thing you could add would be some references as it is good to have a direct comparison between a given equation in a given paper/book and the code implemented.
It appears there are some minor conflicts with the develop branch. This should be easy enough to fix if you pull the latest commits from develop.

Once these points are resolved I will make a final pass, in particular for the moved algorithms, which I believe should not have changed much.

paquiteau · 2021-12-09T09:23:36Z

Thanks for the review.
I added the references where it was required, and documented what was not :)

Also, I change the update_grad_dir and _update_grad_speed to private methods, these should not be changed/called by the user, as they are directly called inside _update.

sfarrens · 2021-12-09T13:36:42Z

Hi @paquiteau great! I still see a couple or merge conflicts listed between your branch and develop.

Conflicting files
modopt/opt/algorithms.py
modopt/tests/test_opt.py

sfarrens · 2021-12-10T09:10:02Z

@paquiteau I have finished checking the other modules and, apart from some very minor things that I can clean up myself when I make the next release, everything looks good.

So, as soon as you resolve the last thread and we figure out the the conflict issue this PR can be merged! 👏

* Add support for tensorflow backend which allows for differentiability (#112) * Added support for tensorflow * Updates to get tests passing * Or --> And * Moving modopt to allow working with tensorflow * Fix issues with wos * Fix all flakes finally! * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Minute updates to codes * Add dynamic module * Fix docu * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Fix 115 (#116) * Fix issues * Add right tests * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Minor bug fix, remove elif (#124) Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Add tests for modopt.base.backend and fix minute bug uncovered (#126) * Minor bug fix, remove elif * Add tests for backend * Fix tests * Add tests * Remove cupy * PEP fixes * Fix PEP * Fix PEP and update * Final PEP * Update setup.cfg Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update test_base.py Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Release cleanup (#128) * updated GPU dependencies * added logo to manifest * updated package version and release date * Unpin package dependencies (#189) * unpinned dependencies * updated pinned documentation dependency versions * Add Gradient descent algorithms (#196) * Version 1.5.1 patch release (#114) * Add support for tensorflow backend which allows for differentiability (#112) * Added support for tensorflow * Updates to get tests passing * Or --> And * Moving modopt to allow working with tensorflow * Fix issues with wos * Fix all flakes finally! * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Minute updates to codes * Add dynamic module * Fix docu * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Fix 115 (#116) * Fix issues * Add right tests * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Minor bug fix, remove elif (#124) Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Add tests for modopt.base.backend and fix minute bug uncovered (#126) * Minor bug fix, remove elif * Add tests for backend * Fix tests * Add tests * Remove cupy * PEP fixes * Fix PEP * Fix PEP and update * Final PEP * Update setup.cfg Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update test_base.py Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Release cleanup (#128) * updated GPU dependencies * added logo to manifest * updated package version and release date Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * make algorithms a module. * add Gradient Descent Algorithms * enforce WPS compliance. * add test for gradient descent * Docstrings improvements * Add See Also and minor corrections * add idx initialisation for all algorithms. * fix merge error * fix typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Release cleanup (#198) * started clean up for next release * update progress * further clean up * additional clean up * cleaned up link to logo * fixed index.rst Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Pierre-Antoine Comby <77174042+paquiteau@users.noreply.github.com>

* Add support for tensorflow backend which allows for differentiability (#112) * Added support for tensorflow * Updates to get tests passing * Or --> And * Moving modopt to allow working with tensorflow * Fix issues with wos * Fix all flakes finally! * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Minute updates to codes * Add dynamic module * Fix docu * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Fix 115 (#116) * Fix issues * Add right tests * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Minor bug fix, remove elif (#124) Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Add tests for modopt.base.backend and fix minute bug uncovered (#126) * Minor bug fix, remove elif * Add tests for backend * Fix tests * Add tests * Remove cupy * PEP fixes * Fix PEP * Fix PEP and update * Final PEP * Update setup.cfg Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update test_base.py Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Release cleanup (#128) * updated GPU dependencies * added logo to manifest * updated package version and release date * Unpin package dependencies (#189) * unpinned dependencies * updated pinned documentation dependency versions * Add Gradient descent algorithms (#196) * Version 1.5.1 patch release (#114) * Add support for tensorflow backend which allows for differentiability (#112) * Added support for tensorflow * Updates to get tests passing * Or --> And * Moving modopt to allow working with tensorflow * Fix issues with wos * Fix all flakes finally! * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Minute updates to codes * Add dynamic module * Fix docu * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Fix 115 (#116) * Fix issues * Add right tests * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Minor bug fix, remove elif (#124) Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Add tests for modopt.base.backend and fix minute bug uncovered (#126) * Minor bug fix, remove elif * Add tests for backend * Fix tests * Add tests * Remove cupy * PEP fixes * Fix PEP * Fix PEP and update * Final PEP * Update setup.cfg Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update test_base.py Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Release cleanup (#128) * updated GPU dependencies * added logo to manifest * updated package version and release date Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * make algorithms a module. * add Gradient Descent Algorithms * enforce WPS compliance. * add test for gradient descent * Docstrings improvements * Add See Also and minor corrections * add idx initialisation for all algorithms. * fix merge error * fix typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Release cleanup (#198) * started clean up for next release * update progress * further clean up * additional clean up * cleaned up link to logo * fixed index.rst * fixed conflict * Fast Singular Value Thresholding (#209) * add SingularValueThreshold This Method provides 10x faster SVT estimation than the LowRankMatrix Operator. * linting * add test for fast computation. * flake8 compliance * Ignore DAR000 Error. * Update modopt/signal/svd.py tuples in docstring Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/signal/svd.py typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/opt/proximity.py typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * update docstring * fix isort * Update modopt/signal/svd.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/signal/svd.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * run isort Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * added writeable input data array feature for benchopt (#213) * removed flake8 limit Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Pierre-Antoine Comby <77174042+paquiteau@users.noreply.github.com>

* Add support for tensorflow backend which allows for differentiability (#112) * Added support for tensorflow * Updates to get tests passing * Or --> And * Moving modopt to allow working with tensorflow * Fix issues with wos * Fix all flakes finally! * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Minute updates to codes * Add dynamic module * Fix docu * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Fix 115 (#116) * Fix issues * Add right tests * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Minor bug fix, remove elif (#124) Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Add tests for modopt.base.backend and fix minute bug uncovered (#126) * Minor bug fix, remove elif * Add tests for backend * Fix tests * Add tests * Remove cupy * PEP fixes * Fix PEP * Fix PEP and update * Final PEP * Update setup.cfg Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update test_base.py Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Release cleanup (#128) * updated GPU dependencies * added logo to manifest * updated package version and release date * Unpin package dependencies (#189) * unpinned dependencies * updated pinned documentation dependency versions * Add Gradient descent algorithms (#196) * Version 1.5.1 patch release (#114) * Add support for tensorflow backend which allows for differentiability (#112) * Added support for tensorflow * Updates to get tests passing * Or --> And * Moving modopt to allow working with tensorflow * Fix issues with wos * Fix all flakes finally! * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Minute updates to codes * Add dynamic module * Fix docu * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Fix 115 (#116) * Fix issues * Add right tests * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Minor bug fix, remove elif (#124) Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Add tests for modopt.base.backend and fix minute bug uncovered (#126) * Minor bug fix, remove elif * Add tests for backend * Fix tests * Add tests * Remove cupy * PEP fixes * Fix PEP * Fix PEP and update * Final PEP * Update setup.cfg Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update test_base.py Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Release cleanup (#128) * updated GPU dependencies * added logo to manifest * updated package version and release date Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * make algorithms a module. * add Gradient Descent Algorithms * enforce WPS compliance. * add test for gradient descent * Docstrings improvements * Add See Also and minor corrections * add idx initialisation for all algorithms. * fix merge error * fix typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Release cleanup (#198) * started clean up for next release * update progress * further clean up * additional clean up * cleaned up link to logo * fixed index.rst * fixed conflict * Fast Singular Value Thresholding (#209) * add SingularValueThreshold This Method provides 10x faster SVT estimation than the LowRankMatrix Operator. * linting * add test for fast computation. * flake8 compliance * Ignore DAR000 Error. * Update modopt/signal/svd.py tuples in docstring Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/signal/svd.py typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/opt/proximity.py typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * update docstring * fix isort * Update modopt/signal/svd.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/signal/svd.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * run isort Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * added writeable input data array feature for benchopt (#213) * removed flake8 limit * updated patch version Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Pierre-Antoine Comby <77174042+paquiteau@users.noreply.github.com>

* Add support for tensorflow backend which allows for differentiability (#112) * Added support for tensorflow * Updates to get tests passing * Or --> And * Moving modopt to allow working with tensorflow * Fix issues with wos * Fix all flakes finally! * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Minute updates to codes * Add dynamic module * Fix docu * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Fix 115 (#116) * Fix issues * Add right tests * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Minor bug fix, remove elif (#124) Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Add tests for modopt.base.backend and fix minute bug uncovered (#126) * Minor bug fix, remove elif * Add tests for backend * Fix tests * Add tests * Remove cupy * PEP fixes * Fix PEP * Fix PEP and update * Final PEP * Update setup.cfg Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update test_base.py Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Release cleanup (#128) * updated GPU dependencies * added logo to manifest * updated package version and release date * Unpin package dependencies (#189) * unpinned dependencies * updated pinned documentation dependency versions * Add Gradient descent algorithms (#196) * Version 1.5.1 patch release (#114) * Add support for tensorflow backend which allows for differentiability (#112) * Added support for tensorflow * Updates to get tests passing * Or --> And * Moving modopt to allow working with tensorflow * Fix issues with wos * Fix all flakes finally! * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/base/backend.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Minute updates to codes * Add dynamic module * Fix docu * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Fix 115 (#116) * Fix issues * Add right tests * Fix PEP Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Minor bug fix, remove elif (#124) Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Add tests for modopt.base.backend and fix minute bug uncovered (#126) * Minor bug fix, remove elif * Add tests for backend * Fix tests * Add tests * Remove cupy * PEP fixes * Fix PEP * Fix PEP and update * Final PEP * Update setup.cfg Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update test_base.py Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Release cleanup (#128) * updated GPU dependencies * added logo to manifest * updated package version and release date Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * make algorithms a module. * add Gradient Descent Algorithms * enforce WPS compliance. * add test for gradient descent * Docstrings improvements * Add See Also and minor corrections * add idx initialisation for all algorithms. * fix merge error * fix typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> Co-authored-by: Chaithya G R <chaithyagr@gmail.com> Co-authored-by: chaithyagr <chaithyagr@gitlab.com> * Release cleanup (#198) * started clean up for next release * update progress * further clean up * additional clean up * cleaned up link to logo * fixed index.rst * fixed conflict * Fast Singular Value Thresholding (#209) * add SingularValueThreshold This Method provides 10x faster SVT estimation than the LowRankMatrix Operator. * linting * add test for fast computation. * flake8 compliance * Ignore DAR000 Error. * Update modopt/signal/svd.py tuples in docstring Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/signal/svd.py typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/opt/proximity.py typo Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * update docstring * fix isort * Update modopt/signal/svd.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/signal/svd.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * run isort Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * added writeable input data array feature for benchopt (#213) * removed flake8 limit * updated patch version * [lint] pydocstyle compliance. (#228) * [lint] pydocstyle compliance. * use pytest-pydocstyle * Power method: fix #211 (#212) * Correct the norm update for Power Method x_new should be divided by its norm, not by x_old_norm. * fix test value We are testing for eigen value of Identity. It should be one. * fix WPS350 * fix test value for unconverged case Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Switch from progressbar to tqdm (#231) * switch from progressbar to tqdm. The progress bar can be provided externally for nested usage. * exposes the progress bar argument. * Child classes better have to implement these. (my linter was complaining) * update docs for progress bar using tqdm. * fix WPS errors * drop progressbar requirement, add tqdm. * [lint] disable warning for non implemented function. * simplify progbar check and argument passthrough * Update README for tqdm dependency (#240) Remote progressbar, use tqdm. * add small help for the metric argument. (#241) * add small help for the metric argument. * RST validation * use single quote * use double backticks. Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * add implementation for admm and fast admm. Based on Goldstein2014 * add Goldstein ref. * WPS compliance. * Abstract class for cost function. * add custom cost operator for admm. * fix WPS compliance. * Ci update (#268) * update python version support. * use string for CI. * remove flake8 and wemake-python-styleguide This anticipates the change to black formatting. * remove wps checks * apparently conda does not support 3.11 for now * remove all linting testing. * fix np.int warning/error * fix dtype error * fix precision for doctest * added black and isort support * Update python version in README * add 3.7 for test back * don't test 3.10 twice * Test rewrite (#266) * add MatrixOperator. * move base test to pytest. * [fixme] remove flake8 and emoji config. * rewrite test_math module using pytest. * use fail/skipparam helper function. * generalize usage of failparam * refactor test_signal. * refactor test_signal, the end. * lint * fix missing parameter. * add dummy object test helper. * rewrite test for cost and gradients. * show missing lines in coverage reports * rewrite of proximity operators testing. * add fail low rank method. * add cases for algorithms test * add algorithm test. * add pytest-cases and pytest-xdists support. * add support for testing metrics. * improve base module coverage. * test for wrong mask in metric module. * add docstring. * update email adress and authors field. * 100% coverage for transform module. * move linear operator to class * update docstring. * paramet(e)rization. * update docstring. * improve test_helper module. * raises should be specified for each failparam call. * encapsulate module's test in classes. * skip test if sklearn is not installed. * pin pydocstyle * removed unnormalised Gaussian kernel option and corresponding test * Restrict scikit-image version for testing * added fix for basic test suite * set behaviour for different astropy versions * updated docstring for gaussian_kernel * Use example scripts as tests. (#277) * Initialize the example module. * do not export the assert statements. * add matplotlib as requirement. * add support for sphinx-gallery * Update modopt/examples/README.rst Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/examples/__init__.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/examples/conftest.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/examples/example_lasso_forward_backward.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * Update modopt/examples/example_lasso_forward_backward.py Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * ignore auto_example folder * doc formatting. * add pogm and basic comparison. * fix: add matplotlib for the plotting in examples scripts. * fix: add matplotlib for basic ci too. * ci: run pytest with xdist for faster testing --------- Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> * fix: specify data_range for ssim. Refs: #290 * typos. * feat(test): add test for admm. * feat(admm): improve doc. * refactor: rename abstract cost to CostParent. * feat: add test for fast admm. * feat(admm): improve docstrings. * style: remove extra line.c * feat: make POGM more memory efficient. * feat: add a dummy cost for the identity operator. * feat: create a linear operator module, add wavelet transform. * feat: add test case for wavelet transform. * Update setup.py --------- Co-authored-by: chaithyagr <chaithyagr@gitlab.com> Co-authored-by: Samuel Farrens <samuel.farrens@gmail.com> Co-authored-by: Pierre-Antoine Comby <77174042+paquiteau@users.noreply.github.com> Co-authored-by: Pierre-antoine Comby <pierre-antoine.comby@crans.org> Co-authored-by: Pierre-Antoine Comby <pierre-antoine.comby@ens-paris-saclay.fr>

sfarrens and others added 3 commits April 22, 2021 09:37

make algorithms a module.

6dfae5f

add Gradient Descent Algorithms

b7fe44a

sfarrens linked an issue Dec 2, 2021 that may be closed by this pull request

[NEW FEATURE] Gradient Descent Algorithms and accelerations methods #194

Closed

2 tasks

sfarrens self-requested a review December 2, 2021 10:41

sfarrens assigned paquiteau Dec 2, 2021

sfarrens added the enhancement label Dec 2, 2021

sfarrens changed the base branch from master to develop December 2, 2021 10:42

enforce WPS compliance.

6cc1b5b

add test for gradient descent

bad50bf

sfarrens requested changes Dec 8, 2021

View reviewed changes

sfarrens reviewed Dec 8, 2021

View reviewed changes

modopt/tests/test_opt.py Show resolved Hide resolved

paquiteau added 2 commits December 9, 2021 10:06

Docstrings improvements

011413b

Add See Also and minor corrections

bd5b4d2

paquiteau requested a review from sfarrens December 9, 2021 13:26

add idx initialisation for all algorithms.

77495f3

paquiteau added 3 commits December 10, 2021 10:18

Merge branch 'develop' into gradient-descent

ecaa9d1

fix merge error

f612ecb

fix typo

d3eb4a0

sfarrens approved these changes Dec 10, 2021

View reviewed changes

sfarrens merged commit 353ef4d into CEA-COSMIC:develop Dec 10, 2021

sfarrens mentioned this pull request Dec 17, 2021

Version 1.6.0 release #200

Merged

paquiteau deleted the gradient-descent branch January 21, 2022 08:48

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Comments

Add Gradient descent algorithms#196

Add Gradient descent algorithms#196
sfarrens merged 11 commits intoCEA-COSMIC:developfrom
paquiteau:gradient-descent

paquiteau commented Dec 1, 2021

Uh oh!

sfarrens commented Dec 2, 2021

Uh oh!

paquiteau commented Dec 2, 2021

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sfarrens commented Dec 8, 2021

Uh oh!

paquiteau commented Dec 9, 2021

Uh oh!

sfarrens commented Dec 9, 2021

Uh oh!

sfarrens commented Dec 10, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Comments

Conversation

paquiteau commented Dec 1, 2021

Uh oh!

sfarrens commented Dec 2, 2021

Uh oh!

paquiteau commented Dec 2, 2021

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

sfarrens commented Dec 8, 2021

Uh oh!

paquiteau commented Dec 9, 2021

Uh oh!

sfarrens commented Dec 9, 2021

Uh oh!

sfarrens commented Dec 10, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants