The Wasserstein Proximal Gradient Algorithm

Abstract

Wasserstein gradient flows are continuous time dynamics that define curves of steepest descent to minimize an objective function over the space of probability measures (i.e., the Wasserstein space). This objective is typically a divergence w.r.t. a fixed target distribution. In recent years, these continuous time dynamics have been used to study the convergence of machine learning algorithms aiming at approximating a probability distribution. However, the discrete-time behavior of these algorithms might differ from the continuous time dynamics. Besides, although discretized gradient flows have been proposed in the literature, little is known about their minimization power. In this work, we propose a Forward Backward (FB) discretization scheme that can tackle the case where the objective function is the sum of a smooth and a nonsmooth geodesically convex terms. Using techniques from convex optimization and optimal transport, we analyze the FB scheme as a minimization algorithm on the Wasserstein space. More precisely, we show under mild assumptions that the FB scheme has convergence guarantees similar to the proximal gradient algorithm in Euclidean spaces.

References in corpus (15)

• Stochastic Optimization for Large-scale Optimal Transport
• Stein Variational Gradient Descent as Gradient Flow
• Convergence of Langevin MCMC in KL-divergence
• Maximum Mean Discrepancy Gradient Flow
• Is There an Analog of Nesterov Acceleration for MCMC?
• On the geometry of Stein variational gradient descent
• Gradient descent algorithms for Bures-Wasserstein barycenters
• Rapid Convergence of the Unadjusted Langevin Algorithm: Isoperimetry Suffices
• Langevin Monte Carlo and JKO splitting
• Policy Optimization as Wasserstein Gradient Flows
• A Unified Theory of SGD: Variance Reduction, Sampling, Quantization and Coordinate Descent
• Proximal Langevin Algorithm: Rapid Convergence Under Isoperimetry
• Approximate inference with Wasserstein gradient flows
• On Wasserstein Reinforcement Learning and the Fokker-Planck equation
• Stochastic Proximal Langevin Algorithm: Potential Splitting and Nonasymptotic Rates