fugw 0.1.1

This release includes:

• dense and sparse solvers for the FUGW optimal transport problem in which the marginal constraints are formulated with an L2 norm instead of a KL (set divergence='kl' or divergence='l2' when defining a mapping) #14 #29
• a series of features and examples to work with volumetric MRI images (as opposed to surface MRI images) #19
• enabling users to set the initial transport plan before fitting a mapping #27
• util functions to save mappings and load them with / without coefficients (for faster loading) #18
• tracking the contribution of each component of the loss throughout training #16
• tracking the evolution of the fugw loss on validation features throughout training #20
• using training callbacks, useful to track custom metrics during training for instance #21 #23
• many bugfixes

Thank you @6Ulm @pbarbarant and @bthirion !

fugw 0.1.0

First release of fugw 🎉

This package provides 3 solvers to approximate solutions to Fused Unbalanced Gromov-Wasserstein (fugw) problems. Moreover, it offers multiple util functions to help computing surface alignments between functional MRI datasets across multiple individuals, and scale to cortical meshes with 150k points or more. Finally, it comes with a documentation containing examples illustrating how to align big distributions.