Skip to yearly menu bar Skip to main content


Geometric Losses for Distributional Learning

Arthur Mensch · Mathieu Blondel · Gabriel Peyré

Pacific Ballroom #179

Keywords: [ Supervised Learning ] [ Structured Prediction ] [ Convex Optimization ]


Building upon recent advances in entropy-regularized optimal transport, and upon Fenchel duality between measures and continuous functions, we propose a generalization of the logistic loss that incorporates a metric or cost between classes. Unlike previous attempts to use optimal transport distances for learning, our loss results in unconstrained convex objective functions, supports infinite (or very large) class spaces, and naturally defines a geometric generalization of the softmax operator. The geometric properties of this loss make it suitable for predicting sparse and singular distributions, for instance supported on curves or hyper-surfaces. We study the theoretical properties of our loss and showcase its effectiveness on two applications: ordinal regression and drawing generation.

Live content is unavailable. Log in and register to view live content