0.9.5

This new release contains several new features, starting with a novel Gaussian Mixture Model Optimal Transport (GMM-OT) solver to compare GMM while enforcing the transport plan to remain a GMM, that benefits from a closed-form solution making it practical for high-dimensional matching problems. We also extended our general unbalanced OT solvers to support any non-negative reference measure in the regularization terms, before adding the novel translation invariant UOT solver showcasing a higher convergence speed.
We also implemented several new solvers and enhanced existing ones to perform OT across spaces. These include a semi-relaxed FGW barycenter solver, coupled with new initialization heuristics for the inner divergence computation, to perform graph partitioning or dictionary learning. Followed by novel unbalanced FGW and Co-optimal transport solvers to promote robustness to outliers in such matching problems. And we finally updated the implementation of partial GW now supporting asymmetric structures and the KL divergence, while leveraging a new generic conditional gradient solver for partial transport problems enabling significant speed improvements. These latest updates required some modifications to the line search functions of our generic conditional gradient solver, paving the way for future improvements to other GW-based solvers.
Last but not least, we implemented a pre-commit scheme to automatically correct common programming mistakes likely to be made by our future contributors.

This release also contains few bug fixes, concerning the support of any metric in ot.emd_1d / ot.emd2_1d, and the support of any weights in ot.gaussian.

Breaking change

• Custom functions provided as parameter line_search to ot.optim.generic_conditional_gradient must now have the signature line_search(cost, G, deltaG, Mi, cost_G, df_G, **kwargs), adding as input df_G the gradient of the regularizer evaluated at the transport plan G. This change aims at improving speed of solvers having quadratic polynomial functions as regularizer such as the Gromov-Wassertein loss (PR #663).

New features

• New linter based on pre-commit using ruff, codespell and yamllint (PR #681)
• Added feature mass=True for nx.kl_div (PR #654)
• Implemented Gaussian Mixture Model OT ot.gmm (PR #649)
• Added feature semirelaxed_fgw_barycenters and generic FGW-related barycenter updates update_barycenter_structure and update_barycenter_feature (PR #659)
• Added initialization heuristics for sr(F)GW problems via semirelaxed_init_plan, integrated in all sr(F)GW solvers (PR #659)
• Improved ot.plot.plot1D_mat (PR #649)
• Added nx.det (PR #649)
• nx.sqrtm is now broadcastable (takes ..., d, d) inputs (PR #649)
• Restructured ot.unbalanced module (PR #658)
• Added ot.unbalanced.lbfgsb_unbalanced2 and add flexible reference measure c in all unbalanced solvers (PR #658)
• Implemented Fused unbalanced Gromov-Wasserstein and unbalanced Co-Optimal Transport (PR #677)
• Notes before depreciating partial Gromov-Wasserstein function in ot.partial moved to ot.gromov (PR #663)
• Create ot.gromov._partial add new features loss_fun = "kl_loss" and symmetry=False to all solvers while increasing speed + updating adequatly ot.solvers (PR #663)
• Added ot.unbalanced.sinkhorn_unbalanced_translation_invariant (PR #676)

Closed issues

• Fixed ot.gaussian ignoring weights when computing means (PR #649, Issue #648)
• Fixed ot.emd_1d and ot.emd2_1d incorrectly allowing any metric (PR #670, Issue #669)

Full Changelog: 0.9.4...0.9.5

0.9.4

This new release contains several new features and bug fixes. Among the new features
we have novel Quantized FGW solvers that can be used to speed up the computation of the FGW loss on large datasets or to promote a structure on the pairwise matrices. We also updated the continuous entropic mapping to provide efficient out-of-sample continuous mapping thanks to entropic regularization. We also have a new general unbalanced solvers for ot.solve and BFGS solver and illustrative example. Finally we have a new solver for the Low Rank Gromov-Wasserstein that can be used to compute the GW distance between two large scale datasets with a low rank approximation.

From a maintenance point of view, we now have a new option to install optional dependencies with pip install POT[all] and the specific backends or submodules' dependencies may also be installed individually. The pip options are: backend-jax, backend-tf, backend-torch, cvxopt, dr, gnn, plot, all. We also provide with this release support for NumPy 2.0 (the wheels should now be compatible with NumPy 2.0 and below). We also fixed several issues such as gradient sign errors for FGW solvers, empty weights for ot.emd2, and line-search in partial GW. We also split the test/test_gromov.py into test/gromov/ to make the tests more manageable.

New features

• NumPy 2.0 support is added (PR #629)
• New quantized FGW solvers ot.gromov.quantized_fused_gromov_wasserstein, ot.gromov.quantized_fused_gromov_wasserstein_samples and ot.gromov.quantized_fused_gromov_wasserstein_partitioned (PR #603)
• ot.gromov._gw.solve_gromov_linesearch now has an argument to specify if the matrices are symmetric in which case the computation can be done faster (PR #607).
• Continuous entropic mapping (PR #613)
• New general unbalanced solvers for ot.solve and BFGS solver and illustrative example (PR #620)
• Add gradient computation with envelope theorem to sinkhorn solver of ot.solve with grad='envelope' (PR #605).
• Added support for Low rank Gromov-Wasserstein with ot.gromov.lowrank_gromov_wasserstein_samples (PR #614)
• Optional dependencies may now be installed with pip install POT[all] The specific backends or submodules' dependencies may also be installed individually. The pip options are: backend-jax, backend-tf, backend-torch, cvxopt, dr, gnn, all. The installation of the cupy backend should be done with conda.

Closed issues

• Fix gpu compatibility of sr(F)GW solvers when G0 is not None(PR #596)
• Fix doc and example for lowrank sinkhorn (PR #601)
• Fix issue with empty weights for ot.emd2 (PR #606, Issue #534)
• Fix a sign error regarding the gradient of ot.gromov._gw.fused_gromov_wasserstein2 and ot.gromov._gw.gromov_wasserstein2 for the kl loss (PR #610)
• Fix same sign error for sr(F)GW conditional gradient solvers (PR #611)
• Split test/test_gromov.py into test/gromov/ (PR #619)
• Fix (F)GW barycenter functions to support computing barycenter on 1 input + deprecate structures as lists (PR #628)
• Fix line-search in partial GW and change default init to the interior of partial transport plans (PR #602)
• Fix ot.da.sinkhorn_lpl1_mm compatibility with JAX (PR #592)
• Fiw linesearch import error on Scipy 1.14 (PR #642, Issue #641)
• Upgrade supported JAX versions from jax<=0.4.24 to jax<=0.4.30 (PR #643)

New Contributors

• @WilliamBonvini made their first contribution in #595
• @KrzakalaPaul made their first contribution in #607
• @matthewfeickert made their first contribution in #629
• @yikun-baio made their first contribution in #602
• @SarahG-579462 made their first contribution in #627
• @simon-forb made their first contribution in #637

Full Changelog: 0.9.3...0.9.4

0.9.3

Closed issues

• Fixed an issue with cost correction for mismatched labels in ot.da.BaseTransport fit methods. This fix addresses the original issue introduced PR #587 (PR #593)

What's Changed

• tiny typos in doc by @gabrielfougeron in #591
• Fix DA cost correction when cost limit is set to Inf by @kachayev in #593
• [MRG] Release 0.9.3 by @rflamary in #594

New Contributors

• @gabrielfougeron made their first contribution in #591

Full Changelog: 0.9.2...0.9.3

0.9.2

This new release contains several new features and bug fixes. Among the new features
we have a new solver for estimation of nearest Brenier potentials (SSNB) that can be used for OT mapping estimation (on small problems), new Bregman Alternated Projected Gradient solvers for GW and FGW, and new solvers for Bures-Wasserstein barycenters. We also provide a first solver for Low Rank Sinkhorn that will be ussed to provide low rak OT extensions in the next releases. Finally we have a new exact line-search for (F)GW solvers with KL loss that can be used to improve the convergence of the solvers.

We also have a new LazyTensor class that can be used to model OT plans and low rank tensors in large scale OT. This class is used to return the plan for the new wrapper for geomloss Sinkhorn solver on empirical samples that can lead to x10/x100 speedups on CPU or GPU and have a lazy implementation that allows solving very large problems of a few millions samples.

We also have a new API for solving OT problems from empirical samples with ot.solve_sample Finally we have a new API for Gromov-Wasserstein solvers with ot.solve_gromov function that centralizes most of the (F)GW methods with unified notation. Some example of how to use the new API below:

# Generate random data
xs, xt = np.random.randn(100, 2), np.random.randn(50, 2)

# Solve OT problem with empirical samples
sol = ot.solve_sample(xs, xt) # Exact OT betwen smaples with uniform weights
sol = ot.solve_sample(xs, xt, wa, wb) # Exact OT with weights given by user 

sol = ot.solve_sample(xs, xt, reg= 1, metric='euclidean') # sinkhorn with euclidean metric

sol = ot.solve_sample(xs, xt, reg= 1, method='geomloss') # faster sinkhorn solver on CPU/GPU

sol = ot.solve_sample(x,x2, method='factored', rank=10) # compute factored OT

sol = ot.solve_sample(x,x2, method='lowrank', rank=10) # compute lowrank sinkhorn OT

value_bw = ot.solve_sample(xs, xt, method='gaussian').value # Bures-Wasserstein distance

# Solve GW problem 
Cs, Ct = ot.dist(xs, xs), ot.dist(xt, xt) # compute cost matrices
sol = ot.solve_gromov(Cs,Ct) # Exact GW between samples with uniform weights

# Solve FGW problem
M = ot.dist(xs, xt) # compute cost matrix

# Exact FGW between samples with uniform weights
sol = ot.solve_gromov(Cs, Ct, M, loss='KL', alpha=0.7) # FGW with KL data fitting  


# recover solutions objects
P = sol.plan # OT plan
u, v = sol.potentials # dual variables
value = sol.value # OT value

# for GW and FGW
value_linear = sol.value_linear # linear part of the loss
value_quad = sol.value_quad # quadratic part of the loss 

Users are encouraged to use the new API (it is much simpler) but it might still be subjects to small changes before the release of POT 1.0.

We also fixed a number of issues, the most pressing being a problem of GPU memory allocation when pytorch is installed that will not happen now thanks to Lazy initialization of the backends. We now also have the possibility to deactivate some backends using environment which prevents POT from importing them and can lead to large import speedup.

New features

• Added support for Nearest Brenier Potentials (SSNB) (PR #526) + minor fix (PR #535)
• Tweaked get_backend to ignore None inputs (PR #525)
• Callbacks for generalized conditional gradient in ot.da.sinkhorn_l1l2_gl are now vectorized to improve performance (PR #507)
• The linspace method of the backends now has the type_as argument to convert to the same dtype and device. (PR #533)
• The convolutional_barycenter2d and convolutional_barycenter2d_debiased functions now work with different devices.. (PR #533)
• New API for Gromov-Wasserstein solvers with ot.solve_gromov function (PR #536)
• New LP solvers from scipy used by default for LP barycenter (PR #537)
• Update wheels to Python 3.12 and remove old i686 arch that do not have scipy wheels (PR #543)
• Upgraded unbalanced OT solvers for more flexibility (PR #539)
• Add LazyTensor for modeling plans and low rank tensor in large scale OT (PR #544)
• Add exact line-search for gromov_wasserstein and fused_gromov_wasserstein with KL loss (PR #556)
• Add KL loss to all semi-relaxed (Fused) Gromov-Wasserstein solvers (PR #559)
• Further upgraded unbalanced OT solvers for more flexibility and future use (PR #551)
• New API function ot.solve_sample for solving OT problems from empirical samples (PR #563)
• Wrapper for `geomloss`` solver on empirical samples (PR #571)
• Add stop_criterion feature to (un)regularized (f)gw barycenter solvers (PR #578)
• Add fixed_structure and fixed_features to entropic fgw barycenter solver (PR #578)
• Add new BAPG solvers with KL projections for GW and FGW (PR #581)
• Add Bures-Wasserstein barycenter in ot.gaussian and example (PR #582, PR #584)
• Domain adaptation method SinkhornL1l2Transport now supports JAX backend (PR #587)
• Added support for Low-Rank Sinkhorn Factorization (PR #568)

Closed issues

• Fix line search evaluating cost outside of the interpolation range (Issue #502, PR #504)
• Lazily instantiate backends to avoid unnecessary GPU memory pre-allocations on package import (Issue #516, PR #520)
• Handle documentation and warnings when integers are provided to (f)gw solvers based on cg (Issue #530, PR #559)
• Correct independence of fgw_barycenters to init_C and init_X (Issue #547, PR #566)
• Avoid precision change when computing norm using PyTorch backend (Discussion #570, PR #572)
• Create ot/bregman/repository (Issue #567, PR #569)
• Fix matrix feature shape in entropic_fused_gromov_barycenters(Issue #574, PR #573)
• Fix (fused) gromov-wasserstein barycenter solvers to support kl_loss(PR #576)

New Contributors

• @laudavid made their first contribution in #568

Full Changelog: 0.9.1...0.9.2

0.9.1

This new release contains several new features and bug fixes.

New features include a new submodule ot.gnn that contains two new Graph neural network layers (compatible with Pytorch Geometric) for template-based pooling of graphs with an example on graph classification. Related to this, we also now provide FGW and semi relaxed FGW solvers for which the resulting loss is differentiable w.r.t. the parameter alpha. Other contributions on the (F)GW front include a new solver for the Proximal Point algorithm that can be used to solve entropic GW problems (using the parameter solver="PPA"), new solvers for entropic FGW barycenters, novels Sinkhorn-based solvers for entropic semi-relaxed (F)GW, the possibility to provide a warm-start to the solvers, and optional marginal weights of the samples (uniform weights ar used by default). Finally we added in the submodule ot.gaussian and ot.da new loss and mapping estimators for the Gaussian Gromov-Wasserstein that can be used as a fast alternative to GW and estimates linear mappings between unregistered spaces that can potentially have different size (See the update linear mapping example for an illustration).

We also provide a new solver for the Entropic Wasserstein Component Analysis that is a generalization of the celebrated PCA taking into account the local neighborhood of the samples. We also now have a new solver in ot.smooth for the sparsity-constrained OT (last plot) that can be used to find regularized OT plans with sparsity constraints. Finally we have a first multi-marginal solver for regular 1D distributions with a Monge loss (see here).

The documentation and testings have also been updated. We now have nearly 95% code coverage with the tests. The documentation has been updated and some examples have been streamlined to build more quickly and avoid timeout problems with CircleCI. We also added an optional CI on GPU for the master branch and approved PRs that can be used when a GPU runner is online.

Many other bugs and issues have been fixed and we want to thank all the contributors, old and new, who made this release possible. More details below.

New features

• Gaussian Gromov Wasserstein loss and mapping (PR #498)
• Template-based Fused Gromov Wasserstein GNN layer in ot.gnn (PR #488)
• Make alpha parameter in semi-relaxed Fused Gromov Wasserstein differentiable (PR #483)
• Make alpha parameter in Fused Gromov Wasserstein differentiable (PR #463)
• Added the sparsity-constrained OT solver to ot.smooth and added projection_sparse_simplex to ot.utils (PR #459)
• Add tests on GPU for master branch and approved PR (PR #473)
• Add median method to all inherited classes of backend.Backend (PR #472)
• Update tests for macOS and Windows, speedup documentation (PR #484)
• Added Proximal Point algorithm to solve GW problems via a new parameter solver="PPA" in ot.gromov.entropic_gromov_wasserstein + examples (PR #455)
• Added features warmstart and kwargs in ot.gromov.entropic_gromov_wasserstein to respectively perform warmstart on dual potentials and pass parameters to ot.sinkhorn (PR #455)
• Added sinkhorn projection based solvers for FGW ot.gromov.entropic_fused_gromov_wasserstein and entropic FGW barycenters + examples (PR #455)
• Added features warmstartT and kwargs to all CG and entropic (F)GW barycenter solvers (PR #455)
• Added entropic semi-relaxed (Fused) Gromov-Wasserstein solvers in ot.gromov + examples (PR #455)
• Make marginal parameters optional for (F)GW solvers in ._gw, ._bregman and ._semirelaxed (PR #455)
• Add Entropic Wasserstein Component Analysis (ECWA) in ot.dr (PR #486)
• Added feature Efficient Discrete Multi Marginal Optimal Transport Regularization + examples (PR #454)

Closed issues

• Fix gromov conventions (PR #497)
• Fix change in scipy API for cdist (PR #487)
• More permissive check_backend (PR #494)
• Fix circleci-redirector action and codecov (PR #460)
• Fix issues with cuda for ot.binary_search_circle and with gradients for ot.sliced_wasserstein_sphere (PR #457)
• Major documentation cleanup (PR #462, PR #467, PR #475)
• Fix gradients for "Wasserstein2 Minibatch GAN" example (PR #466)
• Faster Bures-Wasserstein distance with NumPy backend (PR #468)
• Fix issue backend for ot.sliced_wasserstein_sphere ot.sliced_wasserstein_sphere_unif (PR #471)
• Fix issue with ot.barycenter_stabilized when used with PyTorch tensors and log=True (PR #474)
• Fix utils.cost_normalization function issue to work with multiple backends (PR #472)
• Fix precision error on marginal sums and (Issue #429, PR #496)

New Contributors

• @kachayev made their first contribution in #462
• @liutianlin0121 made their first contribution in #459
• @francois-rozet made their first contribution in #468
• @framunoz made their first contribution in #472
• @SoniaMaz8 made their first contribution in #483
• @tomMoral made their first contribution in #494
• @x12hengyu made their first contribution in #454

Full Changelog: 0.9.0...0.9.1

0.9.0

This new release contains so many new features and bug fixes since 0.8.2 that we decided to make it a new minor release at 0.9.0.

The release contains many new features. First we did a major update of all Gromov-Wasserstein solvers that brings up to 30% gain in
computation time (see PR #431) and allows the GW solvers to work on non symmetricmatrices. It also brings novel solvers for the veryefficient semi-relaxed GW problem that can be used to find the best re-weighting for one of the distributions. We also now have fast and differentiable solvers for Wasserstein on the circle and sliced Wasserstein on the sphere. We are also very happy to provide new OT barycenter solvers such as the Free support Sinkhorn barycenter and the Generalized Wasserstein barycenter. A new differentiable solver for OT across spaces that provides OT plans between samples and features simultaneously and called Co-Optimal Transport has also been implemented. Finally we began working on OT between Gaussian distributions and now provide differentiable estimation for the Bures-Wasserstein divergence and mappings.

Another important first step toward POT 1.0 is the implementation of a unified API for OT solvers with introduction of the ot.solve function that can solve (depending on parameters) exact, regularized and unbalanced OT and return a new OTResult object. The idea behind this new API is to facilitate exploring different solvers with just a change of parameter and get a more unified API for them. We will keep the old solvers API for power users but it will be the preferred way to solve problems starting from release 1.0.0. We provide below some examples of use for the new function and how to recover different aspects of the solution (OT plan, full loss, linear part of the loss, dual variables) :

#Solve  exact ot
sol = ot.solve(M)

# get the results
G = sol.plan # OT plan
ot_loss = sol.value # OT value (full loss for regularized and unbalanced)
ot_loss_linear = sol.value_linear # OT value for linear term np.sum(sol.plan*M)
alpha, beta = sol.potentials # dual potentials

# direct plan and loss computation
G = ot.solve(M).plan
ot_loss = ot.solve(M).value

# OT exact with marginals a/b
sol2 = ot.solve(M, a, b)

# regularized and unbalanced OT
sol_rkl = ot.solve(M, a, b, reg=1) # KL regularization
sol_rl2 = ot.solve(M, a, b, reg=1, reg_type='L2')
sol_ul2 = ot.solve(M, a, b, unbalanced=10, unbalanced_type='L2')
sol_rkl_ukl = ot.solve(M, a, b, reg=10, unbalanced=10) # KL + KL

The function is fully compatible with backends and will be implemented for different types of distribution support (empirical distributions, grids) and OT problems (Gromov-Wasserstein) in the new releases. This new API is not yet presented in the kickstart part of the documentation as there is a small change that it might change when implementing new solvers but we encourage users to play with it.

Finally, in addition to those many new this release fixes 20 issues (some long standing) and we want to thank all the contributors who made this release so big. More details below.

New features

• Added feature to (Fused) Gromov-Wasserstein solvers herited from ot.optim to support relative and absolute loss variations as stopping criterions (PR #431)
• Added feature to (Fused) Gromov-Wasserstein solvers to handle asymmetric matrices (PR #431)
• Added semi-relaxed (Fused) Gromov-Wasserstein solvers in ot.gromov + examples (PR #431)
• Added the spherical sliced-Wasserstein discrepancy in ot.sliced.sliced_wasserstein_sphere and ot.sliced.sliced_wasserstein_sphere_unif + examples (PR #434)
• Added the Wasserstein distance on the circle in ot.lp.solver_1d.wasserstein_circle (PR #434)
• Added the Wasserstein distance on the circle (for p>=1) in ot.lp.solver_1d.binary_search_circle + examples (PR #434)
• Added the 2-Wasserstein distance on the circle w.r.t a uniform distribution in ot.lp.solver_1d.semidiscrete_wasserstein2_unif_circle (PR #434)
• Added Bures Wasserstein distance in ot.gaussian (PR ##428)
• Added Generalized Wasserstein Barycenter solver + example (PR #372), fixed graphical details on the example (PR #376)
• Added Free Support Sinkhorn Barycenter + example (PR #387)
• New API for OT solver using function ot.solve (PR #388)
• Backend version of ot.partial and ot.smooth (PR #388 and #449)
• Added argument for warmstart of dual potentials in Sinkhorn-based methods in ot.bregman (PR #437)
• Added parameters method in ot.da.SinkhornTransport (PR #440)
• ot.dr now uses the new Pymanopt API and POT is compatible with current
  Pymanopt (PR #443)
• Added CO-Optimal Transport solver + examples (PR #447)
• Remove the redundant nx.abs() at the end of wasserstein_1d() (PR #448)

Closed issues

• Fixed an issue with the documentation gallery sections (PR #395)
• Fixed an issue where sinkhorn divergence did not have a gradients (Issue #393, PR #394)
• Fixed an issue where we could not ask TorchBackend to place a random tensor on GPU
  (Issue #371, PR #373)
• Fixed an issue where Sinkhorn solver assumed a symmetric cost matrix (Issue #374, PR #375)
• Fixed an issue where hitting iteration limits would be reported to stderr by std::cerr regardless of Python's stderr stream status (PR #377)
• Fixed an issue where the metric argument in ot.dist did not allow a callable parameter (Issue #378, PR #379)
• Fixed an issue where the max number of iterations in ot.emd was not allowed to go beyond 2^31 (PR #380)
• Fixed an issue where pointers would overflow in the EMD solver, returning an
  incomplete transport plan above a certain size (slightly above 46k, its square being
  roughly 2^31) (PR #381)
• Error raised when mass mismatch in emd2 (PR #386)
• Fixed an issue where a pytorch example would throw an error if executed on a GPU (Issue #389, PR #391)
• Added a work-around for scipy's bug, where you cannot compute the Hamming distance with a "None" weight attribute. (Issue #400, PR #402)
• Fixed an issue where the doc could not be built due to some changes in matplotlib's API (Issue #403, PR #402)
• Replaced Numpy C Compiler with Setuptools C Compiler due to deprecation issues (Issue #408, PR #409)
• Fixed weak optimal transport docstring (Issue #404, PR #410)
• Fixed error with parameter log=Truefor SinkhornLpl1Transport (Issue #412,
  PR #413)
• Fixed an issue about warn parameter in sinkhorn2 (PR #417)
• Fix an issue where the parameter stopThr in empirical_sinkhorn_divergence was rendered useless by subcalls
  that explicitly specified stopThr=1e-9 (Issue #421, PR #422).
• Fixed a bug breaking an example where we would try to make an array of arrays of different shapes (Issue #424, PR #425)
• Fixed an issue with the documentation gallery section (PR #444)
• Fixed issues with cuda variables for line_search_armijo and entropic_gromov_wasserstein (Issue #445, #PR 446)

New Contributors

• @eloitanguy made their first contribution in #372
• @stanleyjs made their first contribution in #377
• @zdk123 made their first contribution in #379
• @clecoz made their first contribution in #386
• @eddardd made their first contribution in #387
• @decarpentierg made their first contribution in #398
• @tlacombe made their first contribution in #423
• @arincbulgur made their first contribution in #417
• @tgnassou made their first contribution in #428
• @clbonet made their first contribution in #434
• @chaosink made their first contribution in #448
• @antoinecollas made their first contribution in #449

Full Changelog: 0.8.2...0.9.0

0.8.2

This releases introduces several new notable features. The less important but most exiting one being that we now have a logo for the toolbox (color and dark background) :

This logo is generated using with matplotlib and using the solution of an OT problem provided by POT (with ot.emd). Generating the logo can be done with a simple python script also provided in the documentation gallery.

New OT solvers include Weak OT and OT with factored coupling that can be used on large datasets. The Majorization Minimization solvers for non-regularized Unbalanced OT are now also available. We also now provide an implementation of GW and FGW unmixing and dictionary learning. It is now possible to use autodiff to solve entropic an quadratic regularized OT in the dual for full or stochastic optimization thanks to the new functions to compute the dual loss for entropic and quadratic regularized OT and reconstruct the OT plan on part or all of the data. They can be used for instance to solve OT problems with stochastic gradient or for estimating the dual potentials as neural networks.

On the backend front, we now have backend compatible functions and classes in the domain adaptation ot.da and unbalanced OT ot.unbalanced modules. This means that the DA classes can be used on tensors from all compatible backends. The free support Wasserstein barycenter solver is now also backend compatible.

Finally we have worked on the documentation to provide an update of existing examples in the gallery and and several new examples including GW dictionary learning and weak Optimal Transport.

New features

• Remove deprecated ot.gpu submodule (PR #361)
• Update examples in the gallery (PR #359)
• Add stochastic loss and OT plan computation for regularized OT and
  backend examples(PR #360)
• Implementation of factored OT with emd and sinkhorn (PR #358)
• A brand new logo for POT (PR #357)
• Better list of related examples in quick start guide with minigallery (PR #334)
• Add optional log-domain Sinkhorn implementation in WDA to support smaller values
  of the regularization parameter (PR #336)
• Backend implementation for ot.lp.free_support_barycenter (PR #340)
• Add weak OT solver + example (PR #341)
• Add backend support for Domain Adaptation and Unbalanced solvers (PR #343)
• Add (F)GW linear dictionary learning solvers + example (PR #319)
• Add links to related PR and Issues in the doc release page (PR #350)
• Add new minimization-maximization algorithms for solving exact Unbalanced OT + example (PR #362)

Closed issues

• Fix mass gradient of ot.emd2 and ot.gromov_wasserstein2 so that they are
  centered (Issue #364, PR #363)
• Fix bug in instantiating an autograd function ValFunction (Issue #337,
  PR #338)
• Fix POT ABI compatibility with old and new numpy (Issue #346, PR #349)
• Warning when feeding integer cost matrix to EMD solver resulting in an integer transport plan (Issue #345, PR #343)
• Fix bug where gromov_wasserstein2 does not perform backpropagation with CUDA
  tensors (Issue #351, PR #352)

0.8.1.0

This is a bug fix release that will remove the benchmarks module form the installation and correct the documentation generation.

Closed issues

• Bug in documentation generation (tag VS master push, PR #332)
• Remove installation of the benchmarks in global namespace (Issue #331, PR #333)

0.8.1

This release fixes several bugs and introduces two new backends: Cupy and Tensorflow. Note that the tensorflow backend will work only when tensorflow has enabled the Numpy behavior (for transpose that is not by default in tensorflow). We also introduce a simple benchmark on CPU GPU for the sinkhorn solver that will be provided in the backend documentation.

This release also brings a few changes in dependencies and compatibility. First we removed tests for Python 3.6 that will not be updated in the future. Also note that POT now depends on Numpy (>= 1.20) because a recent change in ABI is making the wheels non-compatible with older numpy versions. If you really need an older numpy POT will work with no problems but you will need to build it from source.

As always we want to that the contributors who helped make POT better (and bug free).

New features

• New benchmark for sinkhorn solver on CPU/GPU and between backends (PR #316)
• New tensorflow backend (PR #316)
• New Cupy backend (PR #315)
• Documentation always up-to-date with README, RELEASES, CONTRIBUTING and
  CODE_OF_CONDUCT files (PR #316, PR #322).

Closed issues

• Fix bug in older Numpy ABI (<1.20) (Issue #308, PR #326)
• Fix bug in ot.dist function when non euclidean distance (Issue #305, PR #306)
• Fix gradient scaling for functions using nx.set_gradients (Issue #309, PR #310)
• Fix bug in generalized Conditional gradient solver and SinkhornL1L2 (Issue #311, PR #313)
• Fix log error in gromov_barycenters (Issue #317, PR #3018)

0.8.0

This new stable release introduces several important features.

First we now have an OpenMP compatible exact ot solver in ot.emd. The OpenMP version is used when the parameter numThreads is greater than one and can lead to nice speedups on multi-core machines.

Second we have introduced a backend mechanism that allows to use standard POT function seamlessly on Numpy, Pytorch and Jax arrays. Other backends are coming but right now POT can be used seamlessly for training neural networks in Pytorch. Notably we propose the first differentiable computation of the exact OT loss with ot.emd2 (can be differentiated w.r.t. both cost matrix and sample weights), but also for the classical Sinkhorn loss with ot.sinkhorn2, the Wasserstein distance in 1D with ot.wasserstein_1d, sliced Wasserstein with ot.sliced_wasserstein_distance and Gromov-Wasserstein with ot.gromov_wasserstein2 . Examples of how this new feature can be used are now available in the documentation where the Pytorch backend is used to estimate a minimal Wasserstein estimator, a Generative Network (GAN), for a sliced Wasserstein gradient flow and optimizing the Gromov-Wassersein distance. Note that the Jax backend is still in early development and quite slow at the moment, we strongly recommend for Jax users to use the OTT toolbox when possible. As a result of this new feature, the old ot.gpu submodule is now deprecated since GPU implementations can be done using GPU arrays on the torch backends.

Other novel features include implementation for Sampled Gromov Wasserstein and Pointwise Gromov Wasserstein, Sinkhorn in log space with method='sinkhorn_log', Projection Robust Wasserstein, and deviased Sinkorn barycenters.

This release will also simplify the installation process. We have now a pyproject.toml that defines the build dependency and POT should now build even when cython is not installed yet. Also we now provide pe-compiled wheels for linux aarch64 that is used on Raspberry PI and android phones and for MacOS on ARM processors.

Finally POT was accepted for publication in the Journal of Machine Learning Research (JMLR) open source software track and we ask the POT users to cite this paper from now on. The documentation has been improved in particular by adding a "Why OT?" section to the quick start guide and several new examples illustrating the new features. The documentation now has two version : the stable version https://pythonot.github.io/ corresponding to the last release and the master version https://pythonot.github.io/master that corresponds to the current master branch on GitHub.

As usual, we want to thank all the POT contributors (now 37 people have contributed to the toolbox). But for this release we thank in particular Nathan Cassereau and Kamel Guerda from the AI support team at IDRIS for their support to the development of the
backend and OpenMP implementations.

New features

• OpenMP support for exact OT solvers (PR #260)
• Backend for running POT in numpy/torch + exact solver (PR #249)
• Backend implementation of most functions in ot.bregman (PR #280)
• Backend implementation of most functions in ot.optim (PR #282)
• Backend implementation of most functions in ot.gromov (PR #294, PR #302)
• Test for arrays of different type and device (CPU/GPU) (PR #304, #303)
• Implementation of Sinkhorn in log space with method='sinkhorn_log' (PR #290)
• Implementation of regularization path for L2 Unbalanced OT (PR #274)
• Implementation of Projection Robust Wasserstein (PR #267)
• Implementation of Debiased Sinkhorn Barycenters (PR #291)
• Implementation of Sampled Gromov Wasserstein and Pointwise Gromov Wasserstein
  (PR #275)
• Add pyproject.toml and build POT without installing cython first (PR #293)
• Lazy implementation in log space for sinkhorn on samples (PR #259)
• Documentation cleanup (PR #298)
• Two up-to-date documentations for stable
  release and for master branch.
• Building wheels on ARM for Raspberry PI and smartphones (PR #238)
• Update build wheels to new version and new pythons (PR #236, #253)
• Implementation of sliced Wasserstein distance (Issue #202, PR #203)
• Add minimal build to CI and perform pep8 test separately (PR #210)
• Speedup of tests and return run time (PR #262)
• Add "Why OT" discussion to the documentation (PR #220)
• New introductory example to discrete OT in the documentation (PR #191)
• Add templates for Issues/PR on Github (PR#181)

Closed issues

• Debug Memory leak in GAN example (#254)
• DEbug GPU bug (Issue #284, #287, PR #288)
• set_gradients method for JAX backend (PR #278)
• Quicker GAN example for CircleCI build (PR #258)
• Better formatting in Readme (PR #234)
• Debug CI tests (PR #240, #241, #242)
• Bug in Partial OT solver dummy points (PR #215)
• Bug when Armijo linesearch (Issue #184, #198, #281, PR #189, #199, #286)
• Bug Barycenter Sinkhorn (Issue 134, PR #195)
• Infeasible solution in exact OT (Issues #126,#93, PR #217)
• Doc for SUpport Barycenters (Issue #200, PR #201)
• Fix labels transport in BaseTransport (Issue #207, PR #208)
• Bug in emd_1d, non respected bounds (Issue #169, PR #170)
• Removed Python 2.7 support and update codecov file (PR #178)
• Add normalization for WDA and test it (PR #172, #296)
• Cleanup code for new version of flake8 (PR #176)
• Fixed requirements in setup.py (PR #174)
• Removed specific MacOS flags (PR #175)