Home

Awesome

SinkhornAutoDiff - Python toolbox to integrate optimal transport loss functions using automatic differentiation and Sinkhorn's algorithm

Overview

Python toolbox to compute and differentiate Optimal Transport (OT) distances. It computes the cost using (generalization of) Sinkhorn's algorithm [1], which can in turn be applied:

<img src="img/fig-general.jpg" width="80%" alt="Typical example of usage: Sinkhorn for density fitting" align="middle"/>

Organization

Planed Features

Installation

Current implementations are available using the following automatic-differentiation toolboxes:

References:

[1] Marco Cuturi, Sinkhorn Distances: Lightspeed Computation of Optimal Transport, NIPS 2013

[2] N. Bonneel, G. Peyré, M. Cuturi. Wasserstein Barycentric Coordinates: Histogram Regression Using Optimal Transport. ACM Transactions on Graphics (Proc. SIGGRAPH 2016), 35(4), pp. 71:1–71:10, 2016.

[3] A. Rolet, M. Cuturi, G. Peyré. Fast Dictionary Learning with a Smoothed Wasserstein Loss. In Proc. AISTATS'16, pp. 630–638, 2016.

[4] G. Peyré, M. Cuturi, J. Solomon. Gromov-Wasserstein Averaging of Kernel and Distance Matrices. In Proc. ICML'16, pp. 2664–2672, 2016.

[5] J. Solomon, F. de Goes, G. Peyré, M. Cuturi, A. Butscher, A. Nguyen, T. Du, L. Guibas. Convolutional Wasserstein Distances: Efficient Optimal Transportation on Geometric Domains. ACM Transactions on Graphics (Proc. SIGGRAPH 2015), 34(4), pp. 66:1–66:11, 2015.

[6] M. Cuturi, M. Blondel, Soft-DTW: A Differentiable Loss Function for Time Series, ICML 2017.

[7] L. Chizat, G. Peyré, B. Schmitzer, F-X. Vialard. Scaling Algorithms for Unbalanced Transport Problems, Preprint Arxiv:1607.05816, 2016.

[8] Aude Genevay, Gabriel Peyré, Marco Cuturi, Sinkhorn-AutoDiff: Tractable Wasserstein Learning of Generative Models, Preprint Arxiv:1706.00292, 2017

[9] Francois-Xavier Vialard, Gabriel Peyré, Optimal Transport for Diffeomorphic Registration, Jean Feydy, Benjamin Charlier, MICCAI 2017.

[10] G. Peyré, L. Chizat, F-X. Vialard, J. Solomon. Quantum Optimal Transport for Tensor Field Processing. Preprint Arxiv:1612.08731, 2016.

Collaborators

Links

Copyright (c) 2017 Noria's team