Differential Properties of Sinkhorn Approximation for Learning with Wasserstein Distance
Abstract
Applications of optimal transport have recently gained remarkable attention thanks to the computational advantages of entropic regularization. However, in most situations the Sinkhorn approximation of the Wasserstein distance is replaced by a regularized version that is less accurate but easy to differentiate. In this work we characterize the differential properties of the original Sinkhorn distance, proving that it enjoys the same smoothness as its regularized version and we explicitly provide an efficient algorithm to compute its gradient. We show that this result benefits both theory and applications: on one hand, high order smoothness confers statistical guarantees to learning with Wasserstein approximations. On the other hand, the gradient formula allows us to efficiently solve learning and optimization problems in practice. Promising preliminary experiments complement our analysis.
Fichier principal
arxivG3.pdf (413.75 Ko)
Télécharger le fichier
5wlambda.pdf (7.3 Ko)
Télécharger le fichier
5wtilde.pdf (6.32 Ko)
Télécharger le fichier
barywlambda.pdf (9.93 Ko)
Télécharger le fichier
barywtilde.pdf (6.7 Ko)
Télécharger le fichier
datasetcrop.png (10.45 Ko)
Télécharger le fichier
deltasbary2.pdf (6.83 Ko)
Télécharger le fichier
ellipinput.jpg (111.43 Ko)
Télécharger le fichier
| Origin | Files produced by the author(s) |
|---|
| Origin | Files produced by the author(s) |
|---|
| Origin | Files produced by the author(s) |
|---|