Data

Tools

Indexes

Applications

Experiments

Open Source Science

Recent Releases of ott-jax

ott-jax - 0.5.1

What's Changed

  • Maint/fix tests by @michalk8 in https://github.com/ott-jax/ott/pull/609
  • solve tests for soft-sort / unb. Bures + add xlim/ylim options in plot by @marcocuturi in https://github.com/ott-jax/ott/pull/610
  • Fix CI by @michalk8 in https://github.com/ott-jax/ott/pull/611
  • Add OTCP by @michalk8 in https://github.com/ott-jax/ott/pull/612
  • add Dotp cost and corresponding test by @marcocuturi in https://github.com/ott-jax/ott/pull/615
  • parameterize sampler for OTCP as fn by @marcocuturi in https://github.com/ott-jax/ott/pull/617
  • Adds entropy of coupling computations and instantiates Dotp to LRCGeometry by @marcocuturi in https://github.com/ott-jax/ott/pull/619
  • Fix: replace jnp.product with jnp.prod in grid tutorial by @zakil-02 in https://github.com/ott-jax/ott/pull/622
  • simplify _interpolate for progOT by @zsteve in https://github.com/ott-jax/ott/pull/650
  • OTCP tutorial by @EugeneNdiaye in https://github.com/ott-jax/ott/pull/613
  • Modify FenchelConjugateLBFGS to handle data in higher dimensions by @donlap in https://github.com/ott-jax/ott/pull/652
  • replace jaxopt by optax's l-bfgs, add legendre convenience wrapper by @marcocuturi in https://github.com/ott-jax/ott/pull/655
  • Sharding demo by @marcocuturi in https://github.com/ott-jax/ott/pull/658

New Contributors

  • @zakil-02 made their first contribution in https://github.com/ott-jax/ott/pull/622
  • @zsteve made their first contribution in https://github.com/ott-jax/ott/pull/650
  • @EugeneNdiaye made their first contribution in https://github.com/ott-jax/ott/pull/613
  • @donlap made their first contribution in https://github.com/ott-jax/ott/pull/652

Full Changelog: https://github.com/ott-jax/ott/compare/0.5.0...0.5.1

- Python
Published by marcocuturi 10 months ago

ott-jax - 0.5.0

What's Changed

  • add glossary by @marcocuturi in https://github.com/ott-jax/ott/pull/594
  • introduce sinkdiv simplified API, expand glossary with sinkhorn divergence and LR sinkhorn by @marcocuturi in https://github.com/ott-jax/ott/pull/596
  • propagate sinkdiv to Monge gap implementation by @marcocuturi in https://github.com/ott-jax/ott/pull/597
  • add Hungarian solver from optax by @marcocuturi in https://github.com/ott-jax/ott/pull/598
  • Feature/refactor initializers by @michalk8 in https://github.com/ott-jax/ott/pull/599
  • Fix Wrong implementation of u_t(x|z) in Brownian Bridge by @selmanozleyen in https://github.com/ott-jax/ott/pull/600
  • Fix CI by @michalk8 in https://github.com/ott-jax/ott/pull/601
  • Feature/refactor relative epsilon by @michalk8 in https://github.com/ott-jax/ott/pull/602
  • Remove source/target masks by @michalk8 in https://github.com/ott-jax/ott/pull/603

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.9...0.5.0

- Python
Published by michalk8 over 1 year ago

ott-jax - 0.4.9

What's Changed

  • refactor sinkhorn_divergence by @marcocuturi in https://github.com/ott-jax/ott/pull/577
  • automatic scaling of epsilon using std instead of mean by default. by @marcocuturi in https://github.com/ott-jax/ott/pull/578
  • Feature/refactor was solver by @michalk8 in https://github.com/ott-jax/ott/pull/580
  • Fix dtype promotion in Sinkhorn by @michalk8 in https://github.com/ott-jax/ott/pull/583
  • expose SinkhornDivergenceOutput by @marcocuturi in https://github.com/ott-jax/ott/pull/584
  • Switch to jax.random.key by @michalk8 in https://github.com/ott-jax/ott/pull/586
  • fix default for intermediate quadratic layers in ICNN by @marcocuturi in https://github.com/ott-jax/ott/pull/587
  • Enable 3.13 CI by @michalk8 in https://github.com/ott-jax/ott/pull/589
  • Feature/batched vmap by @michalk8 in https://github.com/ott-jax/ott/pull/588

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.8...0.4.9

- Python
Published by michalk8 over 1 year ago

ott-jax - OTT 0.4.8

What's Changed

  • Fix low-rank convergence criterion by @michalk8 in https://github.com/ott-jax/ott/pull/547
  • Fix/quadratic pytree by @michalk8 in https://github.com/ott-jax/ott/pull/555
  • Fix/gw to lr by @michalk8 in https://github.com/ott-jax/ott/pull/558
  • TI transport map by @michalk8 in https://github.com/ott-jax/ott/pull/559
  • Update tutorial structure by @michalk8 in https://github.com/ott-jax/ott/pull/560
  • Fix Grid.prepare_divergences by @michalk8 in https://github.com/ott-jax/ott/pull/563
  • Raise error when instatiating cost/kernel for a grid by @michalk8 in https://github.com/ott-jax/ott/pull/565
  • correct convergence criterion for GW by @marcocuturi in https://github.com/ott-jax/ott/pull/566
  • Feature/progot by @michalk8 in https://github.com/ott-jax/ott/pull/567
  • Fix/univariate non ti by @michalk8 in https://github.com/ott-jax/ott/pull/569
  • sink div for LR by @marcocuturi in https://github.com/ott-jax/ott/pull/568
  • Add tutorial for unbalanced OT by @nvesseron in https://github.com/ott-jax/ott/pull/291
  • Edits to ProgOT by @pkassraie in https://github.com/ott-jax/ott/pull/570
  • Feature/update progot by @michalk8 in https://github.com/ott-jax/ott/pull/571
  • update mm-sink tutorial and plot by @zoepiran in https://github.com/ott-jax/ott/pull/572
  • Feature/update prox by @michalk8 in https://github.com/ott-jax/ott/pull/575
  • implements sliced W by @marcocuturi in https://github.com/ott-jax/ott/pull/576

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.7...0.4.8

- Python
Published by michalk8 over 1 year ago

ott-jax - OTT 0.4.7

What's Changed

  • add dropout to VelocityField and OTFM/GENOT by @MUCDK in https://github.com/ott-jax/ott/pull/521
  • Update icnn.py by @marcocuturi in https://github.com/ott-jax/ott/pull/524
  • Update Codecov Action to v4 by @michalk8 in https://github.com/ott-jax/ott/pull/523
  • Add h-transform for TI costs by @michalk8 in https://github.com/ott-jax/ott/pull/525
  • Update univariate.py by @marcocuturi in https://github.com/ott-jax/ott/pull/527
  • Fix/h legendre tests by @michalk8 in https://github.com/ott-jax/ott/pull/529
  • Update costs.py by @marcocuturi in https://github.com/ott-jax/ott/pull/532
  • Debiased property change by @michalk8 in https://github.com/ott-jax/ott/pull/536
  • Don't stop grads w.r.t. the parametrized costs by @michalk8 in https://github.com/ott-jax/ott/pull/544
  • univariate dual vectors for arbitrary sizes by @marcocuturi in https://github.com/ott-jax/ott/pull/505
  • Fix low-rank convergence criterion by @michalk8 in https://github.com/ott-jax/ott/pull/547
  • Fixing typos in documentation by @pkassraie in https://github.com/ott-jax/ott/pull/548
  • add MMSinkhorn solver by @marcocuturi in https://github.com/ott-jax/ott/pull/549
  • Force ICNN to adopt default initialization of its own layers by @Algue-Rythme in https://github.com/ott-jax/ott/pull/551
  • Deprecate Python 3.8 by @michalk8 in https://github.com/ott-jax/ott/pull/554
  • New neural OT solver ENOT by @nazarblch in https://github.com/ott-jax/ott/pull/503
  • Refactor regularized TI costs by @michalk8 in https://github.com/ott-jax/ott/pull/553

New Contributors

  • @Algue-Rythme made their first contribution in https://github.com/ott-jax/ott/pull/551
  • @nazarblch made their first contribution in https://github.com/ott-jax/ott/pull/503

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.6...0.4.7

- Python
Published by michalk8 almost 2 years ago

ott-jax - OTT 0.4.6

What's Changed

  • Enable M1 runner by @michalk8 in https://github.com/ott-jax/ott/pull/486
  • Potentials as tuple by @michalk8 in https://github.com/ott-jax/ott/pull/488
  • Remove bool option for scale_cost by @michalk8 in https://github.com/ott-jax/ott/pull/492
  • Use more iterations for LR-GW by @michalk8 in https://github.com/ott-jax/ott/pull/494
  • update sinhorn divergence gradient flow nb by @marcocuturi in https://github.com/ott-jax/ott/pull/497
  • add barycenter operator for (convex) translation invariant costs by @marcocuturi in https://github.com/ott-jax/ott/pull/498
  • Addition of gangbo-mccann map estimators using twist operator by @marcocuturi in https://github.com/ott-jax/ott/pull/500
  • Add sparse Chebyshev approximation by @guillaumehu in https://github.com/ott-jax/ott/pull/502
  • Fix/io callback by @michalk8 in https://github.com/ott-jax/ott/pull/506
  • Maint/third lower bound by @michalk8 in https://github.com/ott-jax/ott/pull/507
  • Remove epsilon scheduler in GW by @michalk8 in https://github.com/ott-jax/ott/pull/508
  • Fix SCOT tutorial imports and data by @michalk8 in https://github.com/ott-jax/ott/pull/511
  • restructuring neural models + addition of OT-FM and GENOT by @MUCDK in https://github.com/ott-jax/ott/pull/468
  • computesparselaplacian gives int32 vs int64 index mismatch by @selmanozleyen in https://github.com/ott-jax/ott/pull/510
  • Update One_Sinkhorn.ipynb by @pkassraie in https://github.com/ott-jax/ott/pull/515
  • Update actions' versions by @michalk8 in https://github.com/ott-jax/ott/pull/516
  • Fix/batched non sym cost by @michalk8 in https://github.com/ott-jax/ott/pull/518
  • bug fix: avoid mixing up linear and quadratic in genot by @soerenab in https://github.com/ott-jax/ott/pull/517

New Contributors

  • @selmanozleyen made their first contribution in https://github.com/ott-jax/ott/pull/510
  • @pkassraie made their first contribution in https://github.com/ott-jax/ott/pull/515
  • @soerenab made their first contribution in https://github.com/ott-jax/ott/pull/517

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.5...0.4.6

- Python
Published by michalk8 about 2 years ago

ott-jax - Neural module refactoring

What's Changed

  • Re-run and refactoring of some docs/tutorials + minor changes by @marcocuturi in https://github.com/ott-jax/ott/pull/431
  • Feature/quadratic solve by @michalk8 in https://github.com/ott-jax/ott/pull/433
  • Fix n_iters by @michalk8 in https://github.com/ott-jax/ott/pull/437
  • Fix/rank2 init by @michalk8 in https://github.com/ott-jax/ott/pull/442
  • nn.dataset.GaussianMixture: minor fix to the docstring and standard deviation parameter name by @bamos in https://github.com/ott-jax/ott/pull/445
  • introduce multivariate cdf / quantiles by @marcocuturi in https://github.com/ott-jax/ott/pull/447
  • Feature/update jax version by @michalk8 in https://github.com/ott-jax/ott/pull/448
  • Histogram Transport Implementation by @Daniel-Packer in https://github.com/ott-jax/ott/pull/444
  • Doc refresh by @marcocuturi in https://github.com/ott-jax/ott/pull/450
  • Fix/jax prng deprecation by @michalk8 in https://github.com/ott-jax/ott/pull/455
  • Added CDF distance to Univariate Solver by @Daniel-Packer in https://github.com/ott-jax/ott/pull/451
  • Use PyPI trusted publishers by @michalk8 in https://github.com/ott-jax/ott/pull/459
  • first lb -> third lb by @Daniel-Packer in https://github.com/ott-jax/ott/pull/462
  • Restructure neural module by @michalk8 in https://github.com/ott-jax/ott/pull/466
  • refactoring UnivariateSolver by @marcocuturi in https://github.com/ott-jax/ott/pull/472
  • Feature/geodesic sinkhorn by @guillaumehu in https://github.com/ott-jax/ott/pull/457
  • Enable 3.12 CI by @michalk8 in https://github.com/ott-jax/ott/pull/475
  • Feature/apply lse mode by @michalk8 in https://github.com/ott-jax/ott/pull/474
  • Fix/neural docs by @michalk8 in https://github.com/ott-jax/ott/pull/473
  • Add quadratic layers and enhance ICNNs, update tutorial by @nvesseron in https://github.com/ott-jax/ott/pull/477
  • Fix heat scaling by @guillaumehu in https://github.com/ott-jax/ott/pull/481
  • Add low-rank kernel geometry by @michalk8 in https://github.com/ott-jax/ott/pull/440

New Contributors

  • @Daniel-Packer made their first contribution in https://github.com/ott-jax/ott/pull/444
  • @guillaumehu made their first contribution in https://github.com/ott-jax/ott/pull/457
  • @nvesseron made their first contribution in https://github.com/ott-jax/ott/pull/477

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.4...0.4.5

- Python
Published by michalk8 over 2 years ago

ott-jax - Unbalanced low-rank Gromov-Wasserstein

What's Changed

  • introduce new norm operator that can differentiate through lambda x: norm(x-x) + lighter tests across the board by @marcocuturi in https://github.com/ott-jax/ott/pull/411
  • Feature/better io callbacks by @michalk8 in https://github.com/ott-jax/ott/pull/413
  • Feature/optional dependencies by @michalk8 in https://github.com/ott-jax/ott/pull/412
  • Update pre-commits by @michalk8 in https://github.com/ott-jax/ott/pull/416
  • Feature/move solve by @michalk8 in https://github.com/ott-jax/ott/pull/418
  • Feature/lineax ridge by @michalk8 in https://github.com/ott-jax/ott/pull/424
  • Feature/ulrgw by @michalk8 in https://github.com/ott-jax/ott/pull/410

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.3...0.4.4

- Python
Published by michalk8 over 2 years ago

ott-jax - Add top-k mask and some minor fixes

What's Changed

  • A few changes to increase Coverage by @marcocuturi in https://github.com/ott-jax/ott/pull/390
  • rename ent_reg_cost in sinkhorn solver by @marcocuturi in https://github.com/ott-jax/ott/pull/391
  • Increase coverage of NN modules by including init in tests by @marcocuturi in https://github.com/ott-jax/ott/pull/392
  • add tests for plots in potentials by @marcocuturi in https://github.com/ott-jax/ott/pull/393
  • test Plot class in tools module to increase coverage by @marcocuturi in https://github.com/ott-jax/ott/pull/394
  • Add soft topk_mask operator, fix ranks docs and tests by @marcocuturi in https://github.com/ott-jax/ott/pull/396
  • Add relative_epsilon option to GromovWasserstein by @michalk8 in https://github.com/ott-jax/ott/pull/355
  • [ENH]: Fix misc typos in soft_sort doc by @pierreablin in https://github.com/ott-jax/ott/pull/397
  • Update docs by @michalk8 in https://github.com/ott-jax/ott/pull/398
  • add ability to add titles on animation plots by @marcocuturi in https://github.com/ott-jax/ott/pull/400
  • Update contributing visibility by @michalk8 in https://github.com/ott-jax/ott/pull/399
  • update nb by @marcocuturi in https://github.com/ott-jax/ott/pull/403
  • Fix not being able to jit the soft quantile by @michalk8 in https://github.com/ott-jax/ott/pull/408
  • Test default_progress_fn, skip types.py by @michalk8 in https://github.com/ott-jax/ott/pull/404

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.2...0.4.3

- Python
Published by michalk8 almost 3 years ago

ott-jax - Unbalanced LR Sinkhorn + New MBO Estimators + minor fixes

What's Changed

  • Add a tuto about sparse monge displacements by @pierreablin in https://github.com/ott-jax/ott/pull/374
  • introduce ent_reg_cost / kl_reg_cost in Sinkhorn to clarify which is returned to the user in SinkhornOutput objects. by @marcocuturi in https://github.com/ott-jax/ott/pull/376
  • more precise pydocs for #376 by @marcocuturi in https://github.com/ott-jax/ott/pull/377
  • Update sinkhorn.py by @marcocuturi in https://github.com/ott-jax/ott/pull/378
  • Compute multiple soft-quantiles in one execution without using vmap by @marcocuturi in https://github.com/ott-jax/ott/pull/382
  • incorporate feedback in #382 by @marcocuturi in https://github.com/ott-jax/ott/pull/385
  • Update soft_sort.py by @marcocuturi in https://github.com/ott-jax/ott/pull/387
  • Add elastic-net type costs with an orthogonal linear transformation of the input before it is fed into the regularizer by @michalk8 in https://github.com/ott-jax/ott/pull/383
  • Add monge gap by @theouscidda6 in https://github.com/ott-jax/ott/pull/361
  • Feature/unbalanced LR sinkhorn by @michalk8 in https://github.com/ott-jax/ott/pull/379
  • changes in definition of Monge gap by @marcocuturi in https://github.com/ott-jax/ott/pull/389
  • fix #375, converged flag when predefined number of iterations. by @marcocuturi in https://github.com/ott-jax/ott/pull/386

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.1...0.4.2

- Python
Published by marcocuturi almost 3 years ago

ott-jax - 0.4.1 Various new features (kernel steps for LR Sinkhon, lineax implicit diff), solvers must now be jitted.

Highlights: - reformatted tutorial NBs - progressbars for Sinkhorn - low-rank Sinkhorn solvers can run in kernel mode - discrete_barycenter is refactored using the Solver/Problem model, rather than a simple function. - soft-DTW added as a cost function. - Sparsity inducing (in displacements) costs, see https://arxiv.org/abs/2302.04065

What's Changed

  • Fix typo: Wasserstein-4 -> Wasserstein-2 by @bamos in https://github.com/ott-jax/ott/pull/249
  • Fix/warnings lint by @michalk8 in https://github.com/ott-jax/ott/pull/260
  • Addresses #229: Copyrights unification by @AWehenkel in https://github.com/ott-jax/ott/pull/261
  • fixed the docs about the implicit_diff argument by @gjhuizing in https://github.com/ott-jax/ott/pull/270
  • Correct package requirements for tests. by @AWehenkel in https://github.com/ott-jax/ott/pull/271
  • Add pre-commit CI cache by @michalk8 in https://github.com/ott-jax/ott/pull/279
  • Fix flake8 D10{1,2,3} errors by @giovp in https://github.com/ott-jax/ott/pull/269
  • Added CostFn and Plot to the documentation by @gjhuizing in https://github.com/ott-jax/ott/pull/288
  • add primal_cost property to output of GW, to compute cost without any regularization. by @marcocuturi in https://github.com/ott-jax/ott/pull/286
  • add ruff and remove redundant pre-commits by @giovp in https://github.com/ott-jax/ott/pull/278
  • Remove matplotlib dependency by @michalk8 in https://github.com/ott-jax/ott/pull/304
  • Fix/epsilon regularization by @michalk8 in https://github.com/ott-jax/ott/pull/310
  • Address #252 - Make soft ranking notebook self contained by @AWehenkel in https://github.com/ott-jax/ott/pull/301
  • Fix small TODOs by @michalk8 in https://github.com/ott-jax/ott/pull/311
  • Add CI tests for upstream JAX by @michalk8 in https://github.com/ott-jax/ott/pull/313
  • Subsample initializer by @JTT94 in https://github.com/ott-jax/ott/pull/309
  • feat: add progress reporting to Sinkhorn by @bosr in https://github.com/ott-jax/ott/pull/228
  • Refactor discrete_barycenter as solver, cf. issue #255 by @marcocuturi in https://github.com/ott-jax/ott/pull/316
  • Make random seeding consistent with PRNG keys everywhere instead of seeds by @othmanesebbouh in https://github.com/ott-jax/ott/pull/290
  • Add Soft-DTW cost function by @michalk8 in https://github.com/ott-jax/ott/pull/318
  • Fix/sinkdiv hessian by @michalk8 in https://github.com/ott-jax/ott/pull/321
  • Enable D* checks by @michalk8 in https://github.com/ott-jax/ott/pull/319
  • Fix spelling by @michalk8 in https://github.com/ott-jax/ott/pull/323
  • Remove ignoring passed params in ICNN by @michalk8 in https://github.com/ott-jax/ott/pull/325
  • Fix spelling by @michalk8 in https://github.com/ott-jax/ott/pull/326
  • neuraldual: fix issues with objective-based amortization and no-fine-tuning by @bamos in https://github.com/ott-jax/ott/pull/327
  • Parameter to make line strength proportional to coupling strength in Plot by @gjhuizing in https://github.com/ott-jax/ott/pull/280
  • Remove sparse option in Graph geometry by @michalk8 in https://github.com/ott-jax/ott/pull/329
  • gh: refactor community files by @SauravMaheshkar in https://github.com/ott-jax/ott/pull/331
  • Have a simpler "getting started" notebook by @pierreablin in https://github.com/ott-jax/ott/pull/296
  • update NBs by @marcocuturi in https://github.com/ott-jax/ott/pull/332
  • Docs/fix minor issues by @michalk8 in https://github.com/ott-jax/ott/pull/334
  • Small fix in soft_sort.quantile docstring by @pierreablin in https://github.com/ott-jax/ott/pull/336
  • Fix/jit inside jit by @michalk8 in https://github.com/ott-jax/ott/pull/335
  • typos by @marcocuturi in https://github.com/ott-jax/ott/pull/339
  • Remove assertions in cost functions by @michalk8 in https://github.com/ott-jax/ott/pull/340
  • Fix EntropicPotentials's test by @michalk8 in https://github.com/ott-jax/ott/pull/343
  • Added tutorial on Sinkhorn divergence by @gjhuizing in https://github.com/ott-jax/ott/pull/282
  • Pin chex version in GPU CI by @michalk8 in https://github.com/ott-jax/ott/pull/346
  • Feature/unscaled gw cost by @michalk8 in https://github.com/ott-jax/ott/pull/348
  • Fix typos in notebook by @michalk8 in https://github.com/ott-jax/ott/pull/350
  • docs/progress bar by @bosr in https://github.com/ott-jax/ott/pull/347
  • Fix/default prng by @michalk8 in https://github.com/ott-jax/ott/pull/353
  • tracking progress of gromov-wasserstein by @bosr in https://github.com/ott-jax/ott/pull/351
  • Store individual costs to in the GW barycenter by @michalk8 in https://github.com/ott-jax/ott/pull/354
  • Update brain notebook by @michalk8 in https://github.com/ott-jax/ott/pull/358
  • add dtype to kernel by @MUCDK in https://github.com/ott-jax/ott/pull/365
  • Neural OT notebook: add more 2d datasets by @bamos in https://github.com/ott-jax/ott/pull/303
  • New kernel lr by @meyerscetbon in https://github.com/ott-jax/ott/pull/324
  • Fine grained control of Bures barycenters by @marcocuturi in https://github.com/ott-jax/ott/pull/366
  • use lineax to solve linear system in implicit diff by @marcocuturi in https://github.com/ott-jax/ott/pull/370
  • Fix .readthedocs.yaml by @michalk8 in https://github.com/ott-jax/ott/pull/372

New Contributors

  • @AWehenkel made their first contribution in https://github.com/ott-jax/ott/pull/261
  • @gjhuizing made their first contribution in https://github.com/ott-jax/ott/pull/270
  • @giovp made their first contribution in https://github.com/ott-jax/ott/pull/269
  • @bosr made their first contribution in https://github.com/ott-jax/ott/pull/228
  • @othmanesebbouh made their first contribution in https://github.com/ott-jax/ott/pull/290
  • @pierreablin made their first contribution in https://github.com/ott-jax/ott/pull/296

Full Changelog: https://github.com/ott-jax/ott/compare/0.4.0...0.4.1

- Python
Published by marcocuturi almost 3 years ago

ott-jax - 0.4.0: GPU tests, re-adding jit, improved NN OT solvers, and misc.

What's Changed

  • Fix badges by @michalk8 in https://github.com/ott-jax/ott/pull/211
  • Feature/faster gw init by @michalk8 in https://github.com/ott-jax/ott/pull/213
  • Fix/device array check by @michalk8 in https://github.com/ott-jax/ott/pull/216
  • Remove support for Python 3.7 by @michalk8 in https://github.com/ott-jax/ott/pull/227
  • chore: add yapf config to pyproject.toml by @SauravMaheshkar in https://github.com/ott-jax/ott/pull/230
  • Remove functional API by @michalk8 in https://github.com/ott-jax/ott/pull/222
  • Docs/notebooks myst by @michalk8 in https://github.com/ott-jax/ott/pull/233
  • [ci skip] Add environment.yml for binder by @michalk8 in https://github.com/ott-jax/ott/pull/236
  • Use lax.cond when computing error in Sinkhorn by @michalk8 in https://github.com/ott-jax/ott/pull/235
  • Fix/debiased entropic map by @michalk8 in https://github.com/ott-jax/ott/pull/238
  • Misc/tox by @michalk8 in https://github.com/ott-jax/ott/pull/239
  • Re-add jit argument by @michalk8 in https://github.com/ott-jax/ott/pull/221
  • Add GPU test runner by @michalk8 in https://github.com/ott-jax/ott/pull/242
  • Remove old Transport interface by @michalk8 in https://github.com/ott-jax/ott/pull/243
  • Remove shape check in LRCGeometry by @michalk8 in https://github.com/ott-jax/ott/pull/247
  • Add regularized costs by @michalk8 in https://github.com/ott-jax/ott/pull/244
  • Check only whether problem can be LR by @michalk8 in https://github.com/ott-jax/ott/pull/248
  • Improvements to the neural dual solver by @bamos in https://github.com/ott-jax/ott/pull/219

New Contributors

  • @SauravMaheshkar made their first contribution in https://github.com/ott-jax/ott/pull/230

Full Changelog: https://github.com/ott-jax/ott/compare/0.3.1...0.4.0

- Python
Published by marcocuturi over 3 years ago

ott-jax - 0.3.1 with new cost features, different code layout, changes in potentials handling. Some solvers in function form deprecated.

What's Changed

  • Fix jax.device_put for potentials by @michalk8 in https://github.com/ott-jax/ott/pull/183
  • Introduce primal_cost and dual_cost for Sinkhorn outputs (only primal for LR) for arbitrary geometries. by @marcocuturi in https://github.com/ott-jax/ott/pull/184
  • Feature/src layout by @michalk8 in https://github.com/ott-jax/ott/pull/188
  • Handle better inf with p parameter in SqPNorm by @marcocuturi in https://github.com/ott-jax/ott/pull/189
  • Feature/graph normalized laplacian by @michalk8 in https://github.com/ott-jax/ott/pull/191
  • Update installation instruction by @michalk8 in https://github.com/ott-jax/ott/pull/195
  • Feature/center potentials by @michalk8 in https://github.com/ott-jax/ott/pull/194
  • Remove jit from solvers by @michalk8 in https://github.com/ott-jax/ott/pull/192
  • Accelerating unbalanced OT by @michalk8 in https://github.com/ott-jax/ott/pull/197
  • Feature/depracate functional api by @michalk8 in https://github.com/ott-jax/ott/pull/204
  • Fixes numerical errors in Bures barycenter, and sqrtm, due to low default precision. by @marcocuturi in https://github.com/ott-jax/ott/pull/205
  • Update docstrings for deprecated functions by @michalk8 in https://github.com/ott-jax/ott/pull/207

Full Changelog: https://github.com/ott-jax/ott/compare/0.3.0...0.3.1

- Python
Published by marcocuturi over 3 years ago

ott-jax - New organization for code; Changes in `costs` for `PointClouds`

The main changes in this version are twofold: - Changes in PointCloud geometry, and more specifically handling of cost function. The power parameter that was used to optionally pass a c(x,y) = CostFn(x,y) ** power is now deprecated. One can add it manually by defining a custom CostFn. To compensate this change, a new class of translation invariant costs (TICost) has been created, from which most costs now inherit, defined as c(x,y) = h(x-y). Additionally, to handle Brenier's theorem, the user has the option of passing on the legendre transform of h, h_legendre. - The core folder was too horizontal, containing various modules. It has been reorganized and split into 3 modules that make more sense, problems (to describe OT problems), solvers (solve them) and initializers (as optional modules to help solvers). The latter two have an arborescence that reflects that in problems.

What's Changed

  • fix point_clouds.ipynb by @marcocuturi in https://github.com/ott-jax/ott/pull/159
  • Deprecate power in PointCloud, introduce TICost and use it to compute Entropic (Brenier) maps. by @marcocuturi in https://github.com/ott-jax/ott/pull/167
  • Misc/project structure by @michalk8 in https://github.com/ott-jax/ott/pull/176

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.11...0.3.0

- Python
Published by marcocuturi over 3 years ago

ott-jax - adaptive momentum by default, Euclidean is now SqEuclidean (W2), Euclidean is for W1 and uses `jnp.linalg.norm` to ensure differentiability

What's Changed

  • default momentum with start=300 by @marcocuturi in https://github.com/ott-jax/ott/pull/155
  • Unbalanced gromov by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/128
  • Euclidean / SqEuclidean and changes in power by @marcocuturi in https://github.com/ott-jax/ott/pull/157

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.10...0.2.11

- Python
Published by marcocuturi over 3 years ago

ott-jax - Fix various bugs and add Meta initializer

What's Changed

  • Make GMM documentation more visible, fix bug in M step of EM algorithm by @geoff-davis in https://github.com/ott-jax/ott/pull/144
  • Feature/initializers as literals by @michalk8 in https://github.com/ott-jax/ott/pull/148
  • Fix fused_penalty and scale_cost in LRGW by @michalk8 in https://github.com/ott-jax/ott/pull/147
  • Add dtype property by @michalk8 in https://github.com/ott-jax/ott/pull/150
  • Meta OT Initializer by @bamos in https://github.com/ott-jax/ott/pull/145
  • fix power=1.0 using abs on cost values when needed by @marcocuturi in https://github.com/ott-jax/ott/pull/153

New Contributors

  • @geoff-davis made their first contribution in https://github.com/ott-jax/ott/pull/144

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.9...0.2.10

- Python
Published by marcocuturi over 3 years ago

ott-jax - 0.2.9: new initializations for Sinkhorn, LR Sinkhorn and LR GW (including k-means++ implementations), entropic map, faster Sinkhorn implicit diff

What's Changed

  • Feature/kmeans++ by @michalk8 in https://github.com/ott-jax/ott/pull/120
  • Feature/ Introduce initialization methods for Sinkhorn by @JTT94 in https://github.com/ott-jax/ott/pull/98
  • Pin sphinx-book-theme>=0.3.3 by @michalk8 in https://github.com/ott-jax/ott/pull/123
  • Additional Docstrings and Comments to NeuralDual-Module by @bunnech in https://github.com/ott-jax/ott/pull/122
  • Allow k-means to be differentiable when using while loop by @michalk8 in https://github.com/ott-jax/ott/pull/130
  • Feature/graph geometry by @michalk8 in https://github.com/ott-jax/ott/pull/126
  • LR Sinkhorn improvements by @meyerscetbon in https://github.com/ott-jax/ott/pull/111
  • Refactor GW initialization by @michalk8 in https://github.com/ott-jax/ott/pull/133
  • Change LR initializer to random by @michalk8 in https://github.com/ott-jax/ott/pull/134
  • Generalized k-means initialization for LR Sinkhorn/Gromov-Wasserstein by @michalk8 in https://github.com/ott-jax/ott/pull/135
  • Switch to one linear system per implicit differentiation by @ersisimou in https://github.com/ott-jax/ott/pull/136
  • Feature/disable warm start by @michalk8 in https://github.com/ott-jax/ott/pull/138
  • Add entropic map by @michalk8 in https://github.com/ott-jax/ott/pull/142

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.8...0.2.9

- Python
Published by marcocuturi over 3 years ago

ott-jax - Extension of barycenters to (Fused) GW, several bug fixes around LR Sinkhorn, parallel (segment) code for Sinkhorn, inits for ICNNs

What's Changed

  • Refactoring problems.py by @marcocuturi in https://github.com/ott-jax/ott/pull/86
  • Feature/gw losses by @michalk8 in https://github.com/ott-jax/ott/pull/88
  • Fix/lr linear apply by @michalk8 in https://github.com/ott-jax/ott/pull/91
  • Addition of ICNN Initialization Schemes by @bunnech in https://github.com/ott-jax/ott/pull/90
  • Fix/point cloud apply cost fn by @michalk8 in https://github.com/ott-jax/ott/pull/93
  • Feature/lr gw apply output by @michalk8 in https://github.com/ott-jax/ott/pull/96
  • Remove legacy online option by @michalk8 in https://github.com/ott-jax/ott/pull/97
  • Feature/fast tests by @michalk8 in https://github.com/ott-jax/ott/pull/101
  • Fix/axis norm by @michalk8 in https://github.com/ott-jax/ott/pull/103
  • Generic LR cost decomposition by @michalk8 in https://github.com/ott-jax/ott/pull/99
  • Remove pytest-memray from test req, use only in CI by @michalk8 in https://github.com/ott-jax/ott/pull/107
  • demo and fixes for barycenters of GMMs by @ersisimou in https://github.com/ott-jax/ott/pull/89
  • Add (F)GW barycenters by @michalk8 in https://github.com/ott-jax/ott/pull/87
  • Feature/notebook tests by @michalk8 in https://github.com/ott-jax/ott/pull/108
  • [ci skip] Fix duplicate bibtex entry and a typo by @michalk8 in https://github.com/ott-jax/ott/pull/109
  • GW different cost ranks by @michalk8 in https://github.com/ott-jax/ott/pull/113
  • segment_sinkhorn by @marcocuturi in https://github.com/ott-jax/ott/pull/114
  • Create CODEOFCONDUCT.md by @marcocuturi in https://github.com/ott-jax/ott/pull/118
  • Fix/padded scaling by @michalk8 in https://github.com/ott-jax/ott/pull/116

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.7...0.2.8

- Python
Published by marcocuturi almost 4 years ago

ott-jax - Fixing continuous barycenter, cleaning up docs and more generally the packaging of OTT.

What's Changed

  • Pre-commit by @michalk8 in https://github.com/ott-jax/ott/pull/71
  • Fix/online gw by @michalk8 in https://github.com/ott-jax/ott/pull/80
  • Feature/improve packaging by @michalk8 in https://github.com/ott-jax/ott/pull/82
  • Ersi gauss bary by @ersisimou in https://github.com/ott-jax/ott/pull/81

New Contributors

  • @ersisimou made their first contribution in https://github.com/ott-jax/ott/pull/81

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.6...0.2.7

- Python
Published by marcocuturi almost 4 years ago

ott-jax - Quite a few bug fixes on jitting / scaling of costs / convergence

What's Changed

  • Fix jit issue with relative epsilon and copying epsilons by @JTT94 in https://github.com/ott-jax/ott/pull/52
  • Bug/scale cost by @MUCDK in https://github.com/ott-jax/ott/pull/54
  • ensure 3 geom objects returned when copying epsilons by @JTT94 in https://github.com/ott-jax/ott/pull/56
  • fixed logic of __call__ of LRSinkhorn to prevent uninterpretable error by @MUCDK in https://github.com/ott-jax/ott/pull/57
  • Project import generated by Copybara. by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/58
  • General cost fn by @JTT94 in https://github.com/ott-jax/ott/pull/61
  • Set converged logic by @zoepiran in https://github.com/ott-jax/ott/pull/62
  • Fix LRSinkhornOutput.apply with batches by @michalk8 in https://github.com/ott-jax/ott/pull/63
  • Fix isallgeoms_lr not returning by @michalk8 in https://github.com/ott-jax/ott/pull/67
  • Feature/quadratic problem scale by @michalk8 in https://github.com/ott-jax/ott/pull/66
  • Fix LR tests, fused_penalty property, tolerances by @michalk8 in https://github.com/ott-jax/ott/pull/68
  • add first version of SCOT demonstration notebook (Gromov-Wasserstein for multi-omics) by @antoinebelloir in https://github.com/ott-jax/ott/pull/64
  • Fix LR Gromov memory by @michalk8 in https://github.com/ott-jax/ott/pull/70
  • Fix scale_cost to LRGeometry, float/max-bound by @michalk8 in https://github.com/ott-jax/ott/pull/72
  • Update README to use Markdown math by @adrhill in https://github.com/ott-jax/ott/pull/75
  • Fix neural dual notebook by @lucaeyring in https://github.com/ott-jax/ott/pull/76

New Contributors

  • @JTT94 made their first contribution in https://github.com/ott-jax/ott/pull/52
  • @MUCDK made their first contribution in https://github.com/ott-jax/ott/pull/54
  • @antoinebelloir made their first contribution in https://github.com/ott-jax/ott/pull/64
  • @adrhill made their first contribution in https://github.com/ott-jax/ott/pull/75
  • @lucaeyring made their first contribution in https://github.com/ott-jax/ott/pull/76

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.5...0.2.6

- Python
Published by marcocuturi about 4 years ago

ott-jax - Bug in not_converged flag that was impacting convergence fixed.

What's Changed

  • not converged logic by @zoepiran in https://github.com/ott-jax/ott/pull/50

New Contributors

  • @zoepiran made their first contribution in https://github.com/ott-jax/ott/pull/50

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.4...0.2.5

- Python
Published by marcocuturi about 4 years ago

ott-jax - Wasserstein (free) barycenters + normalizations + alternative initialization for Sinkhorn LR + neural dual solvers

What's Changed

  • Fixes rank and epsilon arguments for LR-GW (see issue #21) by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/28
  • Add Neural Dual Solver and Utilities by @bunnech in https://github.com/ott-jax/ott/pull/32
  • Migrate away from using JaxTestCase in tests by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/35
  • Adding scaling factor to the cost matrix. by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/38
  • Initialization LRSinkhorn by @meyerscetbon in https://github.com/ott-jax/ott/pull/47
  • Adding max_cost scaling for LR. by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/48
  • Free support Wasserstein barycenter by @marcocuturi in https://github.com/ott-jax/ott/pull/49

New Contributors

  • @bunnech made their first contribution in https://github.com/ott-jax/ott/pull/32
  • @meyerscetbon made their first contribution in https://github.com/ott-jax/ott/pull/39

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.3...0.2.4

- Python
Published by marcocuturi about 4 years ago

ott-jax - Improved online mode, entropic regularization for LR-GW, bug fixes in GW, fixing tests

What's Changed

  • Update direct transport plan computation by @theouscidda6 in https://github.com/ott-jax/ott/pull/5
  • Updating GW notebook to work with v0.2.2, adding unbalanced and fused GW options to Transport. by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/4
  • Initialising linear_convergence and costs in GWState to -1 by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/6
  • README example: transport.Transport -> transport.solve by @bamos in https://github.com/ott-jax/ott/pull/7
  • Fixes bug in transport.solve by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/8
  • Remove custom vdot implementation by @michalk8 in https://github.com/ott-jax/ott/pull/10
  • Fixes imports in tests. by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/11
  • Fixes tests. by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/12
  • Fixes tests. by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/13
  • Fixing tests. by @LaetitiaPapaxanthos in https://github.com/ott-jax/ott/pull/14
  • Entropic regularization for LR-GW by @marcocuturi in https://github.com/ott-jax/ott/pull/19
  • Batch apply_lse_kernel for online=True by @michalk8 in https://github.com/ott-jax/ott/pull/23

New Contributors

  • @theouscidda6 made their first contribution in https://github.com/ott-jax/ott/pull/5
  • @LaetitiaPapaxanthos made their first contribution in https://github.com/ott-jax/ott/pull/4
  • @bamos made their first contribution in https://github.com/ott-jax/ott/pull/7
  • @michalk8 made their first contribution in https://github.com/ott-jax/ott/pull/10

Full Changelog: https://github.com/ott-jax/ott/compare/0.2.2...0.2.3

- Python
Published by marcocuturi over 4 years ago

ott-jax - Minor release: Removing icnn from __init__.

Preventing bug in colab when importing ott, as flax is not a preinstalled package.

- Python
Published by LaetitiaPapaxanthos over 4 years ago

ott-jax - Minor release: Switching back to Python 3.7

Switching back to Python 3.7 to avoid bug with colab.

- Python
Published by LaetitiaPapaxanthos over 4 years ago

ott-jax - Version 2.0

Version 2.0 is out with major changes:

  • Toolbox will now be developped in the ott-jax org's repo, and has moved out of google_research. pip path remains the same, i.e. pip install ott-jax.
  • Introduction of LRCGeometry, i.e. low rank geometries that can apply cost in efficient manner.
  • Rewritten the sinkhorn function, which is now a wrapper around a Sinkhorn solver class, using SinkhornState and SinkhornOutput variables. In this refactoring, momentum and anderson are now objects.
  • Same for gromov_wasserstein which now runs solver GromovWasserstein
  • OT problems are now defined in problems and quad_problems
  • Addition of low-rank solvers, both to solve Sinkhorn (LRSinkhorn) and to solve Gromov-Wasserstein (just need to specify a rank).
  • Unbalanced GW solver
  • Tools repository now holds tools to define and manipulate Gaussians and mixtures of Gaussians.
  • and many more...

- Python
Published by marcocuturi over 4 years ago