RELEASES.md

Releases

Skada v0.4.0

This update brings significant enhancements and new features:

1. New Shallow Methods: MongeAlignment and JCPOT
2. New Deep Methods: CAN, MCC, MDD, SPA, SourceOnly, and TargetOnly models.
3. Scorers: Introduced MixValScorer and improved scorer compatibility with deep models.
4. Subsampling Transformers: Added StratifiedDomainSubsampler and DomainSubsampler.
5. Deep Models: Enhanced batch handling, fixed predict_proba, stabilized MDD loss, and fixed Deep Coral.
6. Docs & Design: Added a contributor guide, new logo, and documentation updates.

What's Changed

• Update README.md with zenodo badge by @rflamary in scikit-adaptation#216
• [MRG] Add multi-domain Monge alignment and JCPOT Target shift method by @rflamary in scikit-adaptation#180
• [MRG] Add a parameter base_criterion to deep models by @tgnassou in scikit-adaptation#217
• [MRG] Add new scorer: MixValScorer by @YanisLalou in scikit-adaptation#221
• [MRG] Fix mixval by @antoinecollas in scikit-adaptation#222
• [MRG] Fix batch issue when generating features + add sample_weight in deep models by @YanisLalou in scikit-adaptation#220
• [MRG] Allow model selection cv to handle nd inputs by @YanisLalou in scikit-adaptation#225
• [MRG] In DEV, reshape features to 2D instead of input by @YanisLalou in scikit-adaptation#226
• [MRG] Add utilities functions to the doc by @antoinecollas in scikit-adaptation#227
• Add new logo! by @tgnassou in scikit-adaptation#223
• Fix ImportanceWeightedScorer compatibility with deep learning models by @YanisLalou in scikit-adaptation#232
• [MRG] fix param for Deepjdot by @tgnassou in scikit-adaptation#234
• [MRG] Add SourceOnly and TargetOnly models by @tgnassou in scikit-adaptation#233
• [MRG] Fix docstring for the regulariation parameter of DA loss by @tgnassou in scikit-adaptation#230
• [MRG] Fix order of feature acquisition for deep module by @tgnassou in scikit-adaptation#235
• [MRG] Add recentering in DeepCoral by @tgnassou in scikit-adaptation#242
• [MRG] Add DomainOnlySampler and DomainOnlyDataloader for SourceOnly ou TargetOnly deep methods by @tgnassou in scikit-adaptation#243
• [MRG] Modify sampler to take the max of the two domains by @tgnassou in scikit-adaptation#241
• Fix: Dev scorer wasn't working with SourceOnly and TargetOnly by @YanisLalou in scikit-adaptation#244
• [MRG] Fix deep coral by @antoinecollas in scikit-adaptation#246
• [MRG] Harmonize fixtures by @antoinecollas in scikit-adaptation#248
• [MRG] Bug fix when None in make_da_pipeline by @antoinecollas in scikit-adaptation#256
• [MRG] Handle edge case Mixvalscorer by @YanisLalou in scikit-adaptation#257
• [MRG] Add CAN Method by @YanisLalou in scikit-adaptation#251
• [MRG] Uncomment MMDTarSReweightAdapter tests by @YanisLalou in scikit-adaptation#260
• [MRG] Enhancements to DomainAwareNet and Scorers to handle allow_source arg by @YanisLalou in scikit-adaptation#258
• [MRG] Subsampling transformer by @rflamary in scikit-adaptation#259
• [MRG] Add MCC method by @tgnassou in scikit-adaptation#250
• [MRG] Fix callback issue in CAN by @YanisLalou in scikit-adaptation#265
• [MRG] fix predict_proba for deep method by @tgnassou in scikit-adaptation#247
• Batchnormfix2 by @antoinedemathelin in scikit-adaptation#266
• [MRG] Handle scalar sample domain by @antoinecollas in scikit-adaptation#267
• [MRG] Add DomainAndLabelStratifiedSubsampleTransformer + Fix DomainStratifiedSubsampleTransformer by @YanisLalou in scikit-adaptation#268
• [MRG] Check if sample_domain have only unique domains indexes in check_*_domain by @apmellot in scikit-adaptation#261
• [MRG] Add epsilon in MCC to prevent log(0) by @YanisLalou in scikit-adaptation#270
• [MRG] Handle edge case for DAN by @YanisLalou in scikit-adaptation#271
• [MRG] Handle edge cases for CAN by @YanisLalou in scikit-adaptation#269
• [MRG] Add MDD method by @ambroiseodt in scikit-adaptation#263
• [MRG] Fix dissimilarities computations of Deep CAN by @antoinecollas in scikit-adaptation#274
• [MRG] Remove redundant centroid computation in spherical k-means by @YanisLalou in scikit-adaptation#275
• [MRG] Fix mdd loss by @antoinecollas in scikit-adaptation#277
• [MRG] Apply label smoothing to stabilize MDD by @antoinecollas in scikit-adaptation#279
• [MRG] do not try to complete when X_source is empty by @antoinecollas in scikit-adaptation#280
• [MRG] Add SPA method by @tgnassou in scikit-adaptation#276
• [MRG] Add contributor guide by @tgnassou in scikit-adaptation#282

Full Changelog: https://github.com/scikit-adaptation/skada/compare/0.3.0...0.4.0