| b3044a8 | CJ Carey | 29 September 2023, 00:13:39 UTC | Bump version to 0.7.0 | 29 September 2023, 00:13:39 UTC |
| e5b06fa | CJ Carey | 28 September 2023, 18:21:14 UTC | Fix sklearn compat issues | 28 September 2023, 18:21:14 UTC |
| 4e89e3d | CJ Carey | 28 September 2023, 13:38:39 UTC | Drop old Python versions from the test matrix | 28 September 2023, 13:38:39 UTC |
| 053779a | CJ Carey | 28 September 2023, 13:35:47 UTC | Skip single-feature test for SDML | 28 September 2023, 13:35:47 UTC |
| 81901ee | CJ Carey | 28 September 2023, 13:27:38 UTC | Fix mahalanobis test | 28 September 2023, 13:27:38 UTC |
| 7102bc5 | CJ Carey | 28 September 2023, 13:20:32 UTC | Resolve some SCML issues The sklearn KMeans class now warns if you don't provide a value for the n_init parameter. I'm setting it to the original default, but we may want to consider setting it to 'auto' in the future. | 28 September 2023, 13:20:32 UTC |
| 8d5059c | CJ Carey | 28 September 2023, 13:09:45 UTC | Fix minor lint issues | 28 September 2023, 13:09:45 UTC |
| 084e1a9 | CJ Carey | 28 September 2023, 13:07:54 UTC | Update Python versions to test | 28 September 2023, 13:07:54 UTC |
| 0acf8b1 | CJ Carey | 28 September 2023, 13:03:44 UTC | Remove deprecated assert_warns_message import This was deprecated and is now removed from sklearn. Thankfully, we don't appear to be using it. | 28 September 2023, 13:03:44 UTC |
| 4a6caea | Aurélien Bellet | 21 June 2022, 11:53:05 UTC | fix docstring (#351) | 21 June 2022, 11:53:05 UTC |
| 17216a7 | Maximiliano Vargas | 21 June 2022, 00:09:16 UTC | Rename variables, proposed by issue #257 (#324) * Rename number_constrains to n_constraints * Renamed num_chunks to n_chunks * LMNN k parameter renamed to n_neighbors * Replaced all 'convergence_threshold' with 'tol' * Fix tests * Fixed more test regarding rename of variable * Warnings for n_constrains * Add all warnings regarding n_constrains * Deprecation warnings for n_chunks * Add deprecation warn to n_neighbors * Add convergence_threshold warnings | 21 June 2022, 00:09:16 UTC |
| 8520418 | Nikolaos Michas | 20 June 2022, 23:53:54 UTC | SCML: Raise ValueError if n_features larger than n_triplets (#350) | 20 June 2022, 23:53:54 UTC |
| d78c720 | CJ Carey | 20 June 2022, 13:55:09 UTC | Update CI to test on Python 3.10 | 20 June 2022, 13:55:09 UTC |
| b8530b2 | CJ Carey | 20 June 2022, 13:52:47 UTC | Support for SLEP010 This requires setting a public `n_features_in_` attribute as part of the fit() logic. For details, see: https://scikit-learn-enhancement-proposals.readthedocs.io/en/latest/slep010/proposal.html | 20 June 2022, 13:52:47 UTC |
| 72b76c8 | Nikolaos Michas | 11 March 2022, 06:00:21 UTC | Update SCML docstring (#347) match actual default value of max_iter | 11 March 2022, 06:00:21 UTC |
| 4e0c444 | Maximiliano Vargas | 17 November 2021, 13:54:36 UTC | [MRG][DOC] Fixes almost all warnings in the docs (#338) * Update API names, unuse depretaed html4 * Fixes a lot of warning. Add Methods doctree * More warnings solved * Fix docs dependencies * New style for Example Code and References * Add all Methods to all classes in docstrings, in alphabetical order * Add MetricTransformer and MahalanobisMixin to auto-docs * Delete unused vars in docs. Use simple quotes * Fix identation * Fix Github CI instead of old Travis CI * References Lists are now numbered * RemoveExample Code body almost everywhere * Removed Methods directive. Kept warnings * Deprecated directive now is red as in sklearn | 17 November 2021, 13:54:36 UTC |
| a797635 | Maximiliano Vargas | 11 November 2021, 05:21:00 UTC | [Minor] GUI for matplotlib to run examples (#323) * Proposal to add pyqt5 as GUI to run examples correctly * Add note about matplotlib gui backend instead | 11 November 2021, 05:21:00 UTC |
| 964f28d | Maximiliano Vargas | 02 November 2021, 16:03:12 UTC | Threshold must be a real number (#322) * Add venv to gitignore * Check if threshold is a real value * Simplified threshold type-check * Follow linter rules * Fix last linter error * Add test to check correct behaviour. Sacrified simplicity for the bool case. * Update test. Stick to custom message. It's bool permissive * Explicit boolean permissive case in test * Changed isinstance for custom ValueError message * TypeError for most input. ValueError for String case. | 02 November 2021, 16:03:12 UTC |
| bdfdb24 | Maximiliano Vargas | 02 November 2021, 14:37:51 UTC | Assigned n_basis to SCML when needed for tests. Catch warn when needed as well. (#341) | 02 November 2021, 14:37:51 UTC |
| 6a4aaea | Maximiliano Vargas | 21 October 2021, 15:36:03 UTC | Fix 7 sources of warnings in the tests (#339) * Fix 7 sources of warnings * Fix indentation * Generalized warnings, as old sklearn throw more warnings * Changed np.any() for any() * Fix identation | 21 October 2021, 15:36:03 UTC |
| aaf8d44 | Maximiliano Vargas | 21 October 2021, 12:05:22 UTC | score_pairs refactor (#333) * Remove 3.9 from compatibility * First draft of refactoring BaseMetricLearner and Mahalanobis Learner * Avoid warning related to score_pairs deprecation in tests of pair_calibraiton * Minor fix * Replaced score_pairs with pair_distance in tests * Replace score_pairs with pair_distance inb docs. * Fix weird commit * Update classifiers to use pair_similarity * Updated rst docs * Fix identation * Update docs of score_pairs, get_metric * Add deprecation Test. Fix identation * Fixed changes requested 1 * Fixed changes requested 2 * Add equivalence test, p_dist == p_score * Fix tests and identation. * Fixed changes requested 3 * Fix identation * Last requested changes * Last small detail | 21 October 2021, 12:05:22 UTC |
| e2c3e92 | Maximiliano Vargas | 18 October 2021, 08:30:21 UTC | Change SCML supervised docs (#337) | 18 October 2021, 08:30:21 UTC |
| 44be909 | Maximiliano Vargas | 12 October 2021, 14:21:04 UTC | Disallow 0 on Triplets predictions (#331) * Remove 3.9 from compatibility * Fix Triplets predict function. Made a test to show the point. * Fix identation * Simplified prediction as suggested * Resolved code review comments * Fix weird commit * Simplified assertion | 12 October 2021, 14:21:04 UTC |
| 7a2a49d | Maximiliano Vargas | 08 October 2021, 13:43:29 UTC | Change min sklearn version to 0.21.3 (#334) | 08 October 2021, 13:43:29 UTC |
| 8571f97 | Maximiliano Vargas | 01 October 2021, 15:20:49 UTC | SCML iris test refactor [Minor] (#332) * Add venv to gitignore * Create yml draft Trying to run tests with github actions * Update yml * Update 2 * Update 3 * YML: Only Python3 versions * Add Codecov to CI * Mirroring actual yml from metric_learning repo * Fix codecov * Fix old scikit learn * Update yml * Remove 3.9 from compatibility * Fixed issue with sklearn 0.20 * Delete comments, and unnecesary workflow_dispatch * SCML test refactor * Commented te test | 01 October 2021, 15:20:49 UTC |
| f548b1e | Yuan Tang | 16 September 2021, 13:50:24 UTC | Tidy up files to reflect the migration to GitHub Actions (#328) Signed-off-by: Yuan Tang <terrytangyuan@gmail.com> | 16 September 2021, 13:50:24 UTC |
| 5427159 | Maximiliano Vargas | 16 September 2021, 11:14:08 UTC | YML file: CI Migration to Github Actions (#327) | 16 September 2021, 11:14:08 UTC |
| 10b6d25 | Maximiliano Vargas | 15 September 2021, 19:58:11 UTC | Addressing LFDA sign indeterminacy (#326) * Add venv to gitignore * Note proposed to address sign indeterminancy | 15 September 2021, 19:58:11 UTC |
| 8583df4 | Maximiliano Vargas | 14 September 2021, 13:12:35 UTC | Fix 5 test regarding make_spd_matrix usage from sklearn (#321) * Add venv to gitignore * Fix 5 test regarding make_spd_matrix usage from sklearn | 14 September 2021, 13:12:35 UTC |
| 137880d | Aurélien Bellet | 26 April 2021, 08:27:27 UTC | fix url issue buttons (#319) | 26 April 2021, 08:27:27 UTC |
| d12bf37 | William de Vazelhes | 22 April 2021, 16:33:17 UTC | [MRG+1] add github issues buttons (#316) * add github issues buttons * add blankline at end of file | 22 April 2021, 16:33:17 UTC |
| 7eef7c6 | William de Vazelhes | 15 April 2021, 16:15:51 UTC | [MRG+2] Update repo to work with both new and old scikit-learn (#313) * Update repo to work with both new and old scikit-learn, and add a travis job to test the old scikit-learn * Add random seed for test * fix str representation for various sklearn versions * fix flake8 error * add more samples to the test for robustness * add comment for additional travis test * change str to dict for simplicity * simplify conditional imports with sklearn_shims file * remove typo added blankline * empty commit Co-authored-by: William de Vazelhes 80055062 <william.de.vazelhes@huawei.com> | 15 April 2021, 16:15:51 UTC |
| 66a12ed | Yuan Tang | 04 August 2020, 01:35:39 UTC | Fix CI build - flake8 E231 (#304) | 04 August 2020, 01:35:39 UTC |
| eb878d6 | CJ Carey | 03 August 2020, 21:03:45 UTC | Resolve a numpy deprecation warning | 03 August 2020, 21:03:45 UTC |
| 617adaa | William de Vazelhes | 03 August 2020, 13:08:33 UTC | [MRG] Fix test for components_from_metric and add tests for _check_sdp_from_eigen (#303) * Fix test for components_from_metric and add tests for _check_sdp_from_eigen * Fix trailing whitespace * Remove unused LinAlgError Co-authored-by: William de Vazelhes 80055062 <william.de.vazelhes@huawei.com> | 03 August 2020, 13:08:33 UTC |
| 86a5208 | William de Vazelhes | 30 July 2020, 14:00:34 UTC | Fix matplotlib agg messages | 30 July 2020, 14:00:34 UTC |
| 730abf1 | Aurélien Bellet | 30 July 2020, 13:11:36 UTC | Update paper ref after JMLR acceptance (#301) * update paper ref * wrong space | 30 July 2020, 13:11:36 UTC |
| 9a6fe8a | William de Vazelhes | 02 July 2020, 12:54:13 UTC | Add the right version number everywhere (#300) | 02 July 2020, 12:54:13 UTC |
| d218d78 | William de Vazelhes | 02 July 2020, 12:13:07 UTC | [MRG] Add python requires plus message for old pips (#299) * Add python requires plus message for old pips * Update install requires for scikit-learn | 02 July 2020, 12:13:07 UTC |
| 3e1af68 | Aurélien Bellet | 01 July 2020, 09:50:48 UTC | fix dependencies doc and add pointer to v0.5.0 for earlier Python versions (#298) | 01 July 2020, 09:50:48 UTC |
| a3435f3 | William de Vazelhes | 01 July 2020, 08:05:31 UTC | Update 2019 to 2020 | 01 July 2020, 08:05:31 UTC |
| 49b811d | William de Vazelhes | 30 June 2020, 18:00:49 UTC | replace +1 by inf (#297) | 30 June 2020, 18:00:49 UTC |
| 899ef47 | William de Vazelhes | 28 June 2020, 18:07:06 UTC | Remove modules.rst added by mistake | 28 June 2020, 18:07:06 UTC |
| 6f783de | William de Vazelhes | 26 June 2020, 15:34:47 UTC | [MRG+2] Update the repo for release (#295) * Get rid of deprecations * Some more modifications * more modifs and fixes * remove flake8 error * some fixes * fix * Fixes * Fixes * Add SCML and change version number * add warning for scml, and fix doc generation * small fix * fix | 26 June 2020, 15:34:47 UTC |
| 43a60c9 | Gabriel Rudloff | 17 June 2020, 20:34:46 UTC | SCML : Sparse Compositional Metric Learning (#278) * scml first commit * add scml to __init__.py * fix in components calculation * remove triplet generator, added in triplets PR * change init&fit interface, faster compute & others * added coments & docstrings, small code changes * typos and added choice of gamma & output_iter * some small improvements * lda tail handling rollback * performance improvement by precomputing rand_ints * small fix in components computation * flake8 fix * SCML_global fit fix & other small changes * Proper use of init vars and unsup bases generation * triplet dataset format & remove_y for triplets * adaptation with dataset format * remove labels for triplets and quadruplets * remove labels * remove labels & old fit random_state asignation * compliant with older numpy versions * small typo and fix order * fix n_basis check * initialize_basis_supervised and some refactoring * proper n_basis handling * scml specific tests * remove small mistake * test user input basis * Changed names and messages and some refactoring * triplets in features form passed to _fit * change indeces handlig and edge case fix * name change and typos * improve test_components_is_2D * Replace triplet_diffs option by better aproach * some comments, docstring and refactoring * fix bad triplet set * flake8 fix * SCML doc first draft * find neighbors for every class only once * improve some docstring and warnings * add sklearn compat test * changes to doc * fix and improve tests * use components_from_metric * change TestSCML to object and parametrize tests * fix test_iris * use model._authorized_basis and other fixes * verbose test * revert sum_where * small n_basis warning instead of error * add test iris on triplet_diffs * test lda & triplet_diffs * improved messages * remove quadruplets and triplets from pipeline test * test big n_features * Correct output iters * output_iter on supervised and improved verbose * flake8 fix * bases generation test comments * change big_n_basis_lda error msg * test generated n_basis and basis shape * add mini batch optimization * correct iter convention * eliminate n_samples = 1000 * batch grad refactored * adagrad adaptive learning * int input checks and tests * flake8 fix * no double division and smaller triplets arrays * minor grammar fixes * minor formatting tweaks Co-authored-by: CJ Carey <perimosocordiae@gmail.com> | 17 June 2020, 20:34:46 UTC |
| c15f1c3 | CJ Carey | 27 May 2020, 13:04:31 UTC | Drop support for python 2 and python 3.5 (#291) * Drop Python2 support * Fix test failures and pyflakes error * fix pep8 issue * Stop testing py3.5 * Update README | 27 May 2020, 13:04:31 UTC |
| 2d5a942 | CJ Carey | 22 May 2020, 18:50:35 UTC | Use scipy's logsumexp function (#290) Fixes gh-289 | 22 May 2020, 18:50:35 UTC |
| 8a02af7 | Aurélien Bellet | 31 March 2020, 16:34:47 UTC | [MRG] Allow installation from conda (#283) * add placeholder sentence * update setup.cfg to include license file to dist * update README * README * README typo * update doc * typo | 31 March 2020, 16:34:47 UTC |
| a748c07 | Aurélien Bellet | 26 March 2020, 12:17:17 UTC | install pytest 4 in travis for python 3.4 (#282) | 26 March 2020, 12:17:17 UTC |
| c378e4b | Aurélien Bellet | 04 March 2020, 16:18:31 UTC | [MRG] Correct a few small doc issues following #280 (#281) * add docstrings for constraints class, pairs and chunks methods * fix missing optional values and descriptions, uniformize * fix indentation problems in docstring and uniformize * fix more small things * cosmit * remove unnecessary line * missing blank line for pep8 | 04 March 2020, 16:18:31 UTC |
| 833e186 | Gabriel Rudloff | 04 March 2020, 16:15:28 UTC | [MRG] Learning on Triplets (#279) * add _TripletsClassifierMixin * added doc * remove redundant code * added tests * triplets added to doc autosumary * rephrasing, added docstring and small changes * small rephrasing * small flake8 fix * Handle low number of neighbors for knn triplets * add tests for knn triplet generation * fixed typos and rephrasing * added more tests for knn triplet construction * sorted triplet & fix test_generate_knntriplets_k * added over the edge knn triplets test * multiple small code refactoring * more refactoring * Fix & test unlabeled handling triplet generation * closer unlabeled point * small clarity enhancement & repmat replacement | 04 March 2020, 16:15:28 UTC |
| 1276040 | Aurélien Bellet | 04 March 2020, 14:19:17 UTC | [MRG] Improve docstrings: add them for Constraints class and methods and fix minor problems (#280) * add docstrings for constraints class, pairs and chunks methods * fix missing optional values and descriptions, uniformize * fix indentation problems in docstring and uniformize | 04 March 2020, 14:19:17 UTC |
| e739239 | Gabriel Rudloff | 04 February 2020, 10:10:21 UTC | Fix covariance initialization when matrix is not invertible (#277) * Fix covariance init when matrix is not invertible * replaced import scipy for only required functions * Change inv for pseudo-inv on custom matrix init * Change from EVD to SVD * Roll back to EVD and pseudo inverse of EVD * Fix non-ASCII char * rephrasing warnings * added tests * more rephrasing * fix test * add test * fixes & adds singular pinv test fron eig * fix tolerance of assert * fix tolerance of assert * fix tolerance of assert * fix random seed * isolate random seed setting | 04 February 2020, 10:10:20 UTC |
| 2380f51 | Robin Vogel | 24 January 2020, 13:11:49 UTC | Corrects the forgotten bits of PR #267 (#269) * maj * maj * corrected PR 267 * trailing whitespace * test calibrate_threshold, test predict * maj * Checks estimator is fitted before set threshold * correct failed tests with MockBadClassifier * remove checks * forgot one * missed one check_is_fitted * sklearn changed the assumptions behind check_is_fitted | 24 January 2020, 13:11:49 UTC |
| 1b40c3b | Gabriel Rudloff | 20 January 2020, 13:35:48 UTC | Changes in documentation. Rephrasing, fixed examples, standarized notation, etc. (#274) * Multiple changes to the documentation. Rephrasing, fixed examples and standarized notation, and others. * Forgot to change one A to L * Replaced broken modindex link for module list * fixed compliance with flake8 * Fixed typos, misplaced example, etc * No new bullet and rectification * remove modules index link * add "respectively" * fix rca examples * fix rca examples again | 20 January 2020, 13:35:48 UTC |
| f48a55d | Aurélien Bellet | 13 January 2020, 13:35:15 UTC | Revert changes in #270 due to revert decision in sklearn (#273) | 13 January 2020, 13:35:15 UTC |
| 7a57b06 | Yuan Tang | 07 January 2020, 15:26:56 UTC | Fix broken link to ITML paper (#271) * Update itml.py | 07 January 2020, 15:26:56 UTC |
| 0370198 | Robin Vogel | 03 January 2020, 09:20:44 UTC | Repairs chunk generation for unknown labels, solves issue #260 (#263) * chunks return a map of index to chunk * maj * maj * remove storing of known labels * typo * no self.num_points * tests for unlabeled, repairs chunk generation * maj * testing diff features * corrected test * diff warning * maj * added parameter bound test | 03 January 2020, 09:20:44 UTC |
| 7819e7c | Aurélien Bellet | 13 December 2019, 10:49:47 UTC | Fix failing tests in last build (#270) * add print to test * fix * fix again * fix again * remove attributes from check_is_fitted * if condition based on python version * add TODO everywhere | 13 December 2019, 10:49:47 UTC |
| 7af910f | Robin Vogel | 04 December 2019, 14:03:23 UTC | More systematic checks that an estimator was fit before using its parameters (#267) * maj * added fit checks * maj * Added checks that the function was fitted. * check the input before if model is fitted * made more sensible checks. * added a test for a threshold * added a test for the unset threshold | 04 December 2019, 14:03:23 UTC |
| 710379e | Robin Vogel | 21 November 2019, 14:01:55 UTC | [DOC] solves several small raised issues (#266) * solves #258, #261, and #256 | 21 November 2019, 14:01:55 UTC |
| 65a98cc | Robin Vogel | 20 November 2019, 15:10:36 UTC | Correct display of references problem in Sphinx 2.1.1 (#265) Reverts to html4 mode because themes haven't updated yet. | 20 November 2019, 15:10:36 UTC |
| 7700529 | Robin Vogel | 13 November 2019, 12:37:36 UTC | Break chunks generation in RCA when not enough possible chunks, fixes issue #200 (#254) * fixes issue 200 * maj * add max_chunks in error message * tests the building of chunks in constraints.py * corrected faulty generation * still small mistake at generation * encapsulate tests, modified message * Testing chunk generation in constraints | 13 November 2019, 12:37:36 UTC |
| b871028 | Robin Vogel | 28 October 2019, 16:10:20 UTC | Solves #225 : adds first line to lmnn's verbose (#253) * solved issue 225 * maj * maj * Delete test.py | 28 October 2019, 16:10:20 UTC |
| a304714 | LBrummer | 12 October 2019, 13:04:42 UTC | Fixed typo in tuple description. (#252) | 12 October 2019, 13:04:42 UTC |
| f65bfd6 | Aurélien Bellet | 14 August 2019, 16:49:47 UTC | ref in doc + fix small typo (#249) | 14 August 2019, 16:49:47 UTC |
| 90b54f2 | Aurélien Bellet | 14 August 2019, 08:21:43 UTC | [MRG] Update README with paper reference (#248) * update readme with paper ref * add bibtex * fix indent * weird appearance * try no indent | 14 August 2019, 08:21:43 UTC |
| 5ad94c3 | Aurélien Bellet | 02 August 2019, 15:08:22 UTC | [MRG] Consistent readme and doc index (#245) * consistent readme and doc index * lower case m | 02 August 2019, 15:08:22 UTC |
| 1e42acb | William de Vazelhes | 02 August 2019, 14:15:26 UTC | Rename repo address (#244) | 02 August 2019, 14:15:26 UTC |
| ca1109b | Yuan Tang | 02 August 2019, 14:14:08 UTC | Extract skggm commit hash to an environment variable (#241) | 02 August 2019, 14:14:08 UTC |
| 66457b5 | Yuan Tang | 02 August 2019, 14:13:04 UTC | Only build on master branch (#242) | 02 August 2019, 14:13:04 UTC |
| 274edea | William de Vazelhes | 01 August 2019, 15:15:59 UTC | Add link to github page in documentation | 01 August 2019, 15:15:59 UTC |
| f67d0e9 | William de Vazelhes | 01 August 2019, 15:11:44 UTC | Update links in documentation | 01 August 2019, 15:11:44 UTC |
| 966a418 | William de Vazelhes | 01 August 2019, 14:50:17 UTC | Update some references in README.rst | 01 August 2019, 14:50:17 UTC |
| 4e57071 | William de Vazelhes | 01 August 2019, 14:30:38 UTC | empty commit to relaunch CI | 01 August 2019, 14:30:38 UTC |
| fa339f8 | William de Vazelhes | 01 August 2019, 09:19:43 UTC | [MRG] Address comments for sklearn-contrib integration (#238) * Fix flake8 errors: Mostly: - blank spaces (missing or to be removed) - lines too long - unused variables or imports - bad indentation * Remove unused pytest import * Update .travis.yml to put also python 3.7 * no more flake8 error * Add flake8 check * Fix flake8 command * Solve W605 errors and update list of ignored errors * Improve flake8 command * Add file to run travis flake8 only on diff * Fix typo * Put good name for flake8 file * Introduce a PEP8 error to see if it works * Introduce another pep8 error, inline this time * Actually call check_files * Be less tolerant on pep8 checks * Test new way to specify travis jobs * Go back to full PEP8 * Remove PEP8 errors introduced for testing * Fix missing entry * Try something with removing 'after_success' * Put an error in test to see what happens * Remove unused flake8 install for pytest runs and rename flake8 job * Add error for SDML, and remove the other error * Remove file * Put again after success * Remove skggm error | 01 August 2019, 09:19:43 UTC |
| c750ef5 | William de Vazelhes | 29 July 2019, 08:54:05 UTC | Remove message about shogun in LMNN (#239) | 29 July 2019, 08:54:05 UTC |
| e7837a1 | Yuan Tang | 19 July 2019, 16:01:36 UTC | Update author information in setup.py (#236) | 19 July 2019, 16:01:36 UTC |
| 57086e9 | William de Vazelhes | 18 July 2019, 14:24:46 UTC | Prepare new version (#232) * Prepare new version * Update authors * reorder authors | 18 July 2019, 14:24:46 UTC |
| 7d74671 | William de Vazelhes | 09 July 2019, 04:28:15 UTC | Fix random state in algorithms (#234) | 09 July 2019, 04:28:15 UTC |
| 54c9d89 | William de Vazelhes | 06 July 2019, 16:22:49 UTC | make check_preprocessor private (#235) | 06 July 2019, 16:22:49 UTC |
| 09dcd56 | William de Vazelhes | 04 July 2019, 07:11:27 UTC | [MRG] update impostors, closer to original implem (#228) * first attempt to change the function * Add test and make it work * Import the right scipy * Add test where the number of impostors varies and tests the gradient * fix little pbs * Fix L_next * Fix LMNN * add forgotten L as argument * add forgotten L as argument * fix cost fn call * fix cost fn call * nitpicks * make example work and fix python2 error | 04 July 2019, 07:11:27 UTC |
| 731b327 | Aurélien Bellet | 04 July 2019, 06:48:28 UTC | [MRG] Deprecate use_pca parameter of LMNN (#231) * deprecate use_pca * fix failing test | 04 July 2019, 06:48:28 UTC |
| a7ede57 | Aurélien Bellet | 04 July 2019, 06:46:51 UTC | Rename transformer_ to components_ (#230) | 04 July 2019, 06:46:51 UTC |
| 44fd427 | William de Vazelhes | 03 July 2019, 22:20:35 UTC | [MRG] Remove random_seed in fit and use the one in init (#224) * Remove random_seed in fit and use the one in init * update tests with the new API * Update test_RCA in sklearn_compat * Update test_SDML in sklearn_compat * Remove testing of pca_comps since it's deprecated * Fix sklearn test * Fix random_seed for test_iris in TestRCA * Relaunch CI * Augment tolerance rather than fix random_seed * Add ChangedBehaviorWarning if the random_state is left default * Update the merge * Address https://github.com/metric-learn/metric-learn/pull/224#pullrequestreview-257618709 | 03 July 2019, 22:20:35 UTC |
| 46a948a | Aurélien Bellet | 03 July 2019, 12:02:33 UTC | [WIP] Small improvements in doc (#229) * get started * supervised.rst done * unsupervised.rst done * finish addressing comments * fix links and uniformize refs | 03 July 2019, 12:02:33 UTC |
| 2dc9b90 | William de Vazelhes | 03 July 2019, 09:06:44 UTC | [MRG] Enhance documentation (#208) * Add link to algorithm in the title of sections * Separate supervised and weakly supervised algorithms in Package Overview * Improve doc * Fix plot_metric_example * Update links with the new pages * Use random init for LMNN * Update remaining num_dims * Add description of the API for Weakly Supervised and Supervised algorithms * Remove to have the list of methods in the description of the class * Fix: Update wronly merged lmnn with None instead of 'auto' in init * Add some documentation about generating pairs and quadruplets * Add base module quick description * Put classes rather than modules * Add docstrings at top of algos * Update name of API details to Package Contents * Update name of API details to Package Contents | 03 July 2019, 09:06:44 UTC |
| 580d38d | William de Vazelhes | 25 June 2019, 14:16:37 UTC | [MRG] fix quadruplets decision_function (#217) * fix quadruplets decision_function * Address https://github.com/metric-learn/metric-learn/pull/217#issuecomment-501792220 * fix: I put the column at the wrong side, now it does some subsampling * Fix number of samples * let's try again with 30 samples * Use less chunks | 25 June 2019, 14:16:37 UTC |
| 8c3cb3e | William de Vazelhes | 18 June 2019, 15:07:53 UTC | [MRG] Fix quadruplets scoring (#220) * FIX: fix lsml scoring * Address https://github.com/metric-learn/metric-learn/pull/220#pullrequestreview-249487025 | 18 June 2019, 15:07:53 UTC |
| 8518517 | William de Vazelhes | 12 June 2019, 15:19:04 UTC | [MRG] Remove preprocessing the data for RCA (#194) * Remove initialization of the data for RCA * Add deprecated flag for supervised version too * Remove comment saying we'll do PCA * Add ChangedBehaviorWarning and do tests * improve change behavior warning * Update message in case covariance matrix is not invertible * FIX: still ignore testing RCA while fixed in #198 * Some reformatting * Fix test string * TST: add test for warning message when covariance is not definite * Address https://github.com/metric-learn/metric-learn/pull/194#discussion_r292387277 | 12 June 2019, 15:19:04 UTC |
| a22c2e6 | William de Vazelhes | 12 June 2019, 14:19:47 UTC | [MRG] Remove shogun dependency (#216) * Remove shogun dependency * Finalize removing of shogun LMNN * Remove LMNN useless base class | 12 June 2019, 14:19:47 UTC |
| 999cb5b | William de Vazelhes | 12 June 2019, 13:27:45 UTC | [MRG] Add ChangedBehaviorWarning message for LMNN too (#214) * Add ChangedBehaviorWarning message for LMNN too * Remove useless 'as an init' | 12 June 2019, 13:27:45 UTC |
| 130cbad | William de Vazelhes | 07 June 2019, 15:26:57 UTC | [MRG] Uniformize initialization for all algorithms (#195) * initiate PR * Revert "initiate PR" This reverts commit a2ae9e10932e58448292e7e3412958649ec3c2d0. * FEAT: uniformize init for NCA and RCA * Let the check of num_dims be done in the other PR * Add metric initialization for algorithms that learn a mahalanobis matrix * Add initialization for MLKR * FIX: fix error message for dimension * FIX fix StringRepr for MLKR * FIX tests by reshaping to the right dataset size * Remove lda in docstring of MLKR * MAINT: Add deprecation for previous initializations * Update tests with new initialization * Make random init for mahalanobis metric generate an SPD matrix * Ensure the input mahalanobis metric initialization is symmetric, and say it should be SPD * various fixes * MAINT: various refactoring - MLKR: update default test init - SDML: refactor prior_inv * FIX fix default covariance for SDML in tests * Enhance docstring * Set random state for SDML * Fix merge remove_spaces that was forgotten * Fix indent * XP: try to change the way we choose n_components to see if it fixes the test * Revert "XP: try to change the way we choose n_components to see if it fixes the test" This reverts commit e86b61b7db2f432d291f8fd85e90ae80b55adf5e. * Be more tolerant in test * Add test for singular covariance matrix * Fix test_singular_covariance_init * DOC: update docstring saying pseudo-inverse * Revert "Fix test_singular_covariance_init" This reverts commit d2cc7cec3099edb1cd1bfaf761e24ec7473e110f. * Ensure definiteness before returning the inverse * wip deal with non definiteness * Rename init to prior for SDML and LSML * Update error messages with either prior or init * Remove message * A few nitpicks * PEP8 errors + change init in test * STY: PEP8 fixes * Address and remove TODOs * Replace init by prior for ITML * TST: fix ITML test with init changed into prior * Add precision for MMC * Add ChangedBehaviorWarning for the algorithms that changed * Address https://github.com/metric-learn/metric-learn/pull/195#pullrequestreview-245440568 * Remove the warnings check since we now have a ChangedBehaviorWarning * Be more precise: it should not raise any ConvergenceWarningError * Address https://github.com/metric-learn/metric-learn/pull/195#pullrequestreview-245911227 * FIX remaining comment * TST: update test error message * Improve readability * Address https://github.com/metric-learn/metric-learn/pull/195#pullrequestreview-246553439 * TST: Fix docsting lmnn * Fix warning messages * Fix warnings messages changed | 07 June 2019, 15:26:57 UTC |
| 3899653 | William de Vazelhes | 07 June 2019, 09:50:13 UTC | [MRG] Uniformize num_dims to n_components and add it for LMNN (#193) * Uniformize num_dims and add it for LMNN * MAINT: fix imports * Fix: fix test_num_dims * MAINT: Address https://github.com/metric-learn/metric-learn/pull/193#pullrequestreview-228280763 * Refactor num_dims in n_components and add deprecation * FIX make some tests work * FIX Make tests work (fix deprecation messages and fix RCA example) * Remove unused import * Revert "Remove unused import" This reverts commit 81c9a8d1a7b4aca1db58153d6dd2b85dceffe0c7. * Fix import * FIX fix some tests * Allow more general sign switching in test_lfda | 07 June 2019, 09:50:13 UTC |
| efba316 | William de Vazelhes | 05 June 2019, 10:40:36 UTC | [MRG] Use pseudo-inverse in Covariance (#206) * FIX: fix covariance algo * some fixes and add non regression test * Use size instead of len * Address https://github.com/metric-learn/metric-learn/pull/206#pullrequestreview-240810281 | 05 June 2019, 10:40:36 UTC |
| fbd92ff | William de Vazelhes | 05 June 2019, 09:44:24 UTC | [MRG] Export notebook to gallery (#180) * Export notebook to gallery * Fix the figure number in order to get the image printed in the logo of the example * wip replace dataset by faces * Finalize notebook * change dataset for make_classification * Add comments on the properties of the algorithms * Address https://github.com/metric-learn/metric-learn/pull/180#pullrequestreview-242334192 * Address https://github.com/metric-learn/metric-learn/pull/180#pullrequestreview-243366233 * a few updates and minor corrections | 05 June 2019, 09:44:24 UTC |
| 187b59e | William de Vazelhes | 29 May 2019, 14:17:02 UTC | [MRG] FIX LMNN gradient and cost function (#201) * TST: make tests for LMNN gradient * FIX: fix gradient computation * Simplify expression * Be more tolerant for checking NCA * Address https://github.com/metric-learn/metric-learn/pull/201#discussion_r286032898 * Add checks for bounds argument * Revert "Add checks for bounds argument" This reverts commit 562f33bfcc7d5fe6a8fc6f65145f4a0d909224d6. * Add missing return | 29 May 2019, 14:17:02 UTC |
| f407bac | William de Vazelhes | 24 May 2019, 17:49:03 UTC | Add checks for bounds argument of ITML (#207) | 24 May 2019, 17:49:03 UTC |
| aa5b274 | William de Vazelhes | 17 May 2019, 07:45:12 UTC | TST: remove comment in test since #175 is fixed | 17 May 2019, 07:45:12 UTC |
| 9f73250 | William de Vazelhes | 13 May 2019, 08:49:56 UTC | [MRG] Fix RCA_Supervised sklearn compat test (#198) * FIX fix RCA_Supervised sklearn compat test * Address https://github.com/metric-learn/metric-learn/pull/198#pullrequestreview-234140017 * Refactor comment | 13 May 2019, 08:49:56 UTC |
