Stochastic Wasserstein Barycenters

Abstract

We present a stochastic algorithm to compute the barycenter of a set of probability distributions under the Wasserstein metric from optimal transport. Unlike previous approaches, our method extends to continuous input distributions and allows the support of the barycenter to be adjusted in each iteration. We tackle the problem without regularization, allowing us to recover a sharp output whose support is contained within the support of the true barycenter. We give examples where our algorithm recovers a more meaningful barycenter than previous work. Our method is versatile and can be extended to applications such as generating super samples from a given distribution and recovering blue noise approximations.

arXiv commentICML 2018

References in corpus (4)

• Fast Computation of Wasserstein Barycenters
• Sliced Wasserstein Kernel for Persistence Diagrams
• A Sparse Multi-Scale Algorithm for Dense Optimal Transport
• Parallel Streaming Wasserstein Barycenters

Cited by in corpus (26)

• Semi-dual Regularized Optimal Transport
• Gradient descent algorithms for Bures-Wasserstein barycenters
• Rates of Estimation of Optimal Transport Maps using Plug-in Estimators via Barycentric Projections
• On the Complexity of Approximating Wasserstein Barycenter
• An Optimal Transport Approach to Personalized Federated Learning
• Decentralize and Randomize: Faster Algorithm for Wasserstein Barycenters
• Scalable Optimal Transport Methods in Machine Learning: A Contemporary Survey
• Scalable Computations of Wasserstein Barycenter via Input Convex Neural Networks
• Scalable multi-class sampling via filtered sliced optimal transport
• Fixed-Support Wasserstein Barycenters: Computational Hardness and Fast Algorithm
• Continuous Regularized Wasserstein Barycenters
• Unsupervised Multilingual Alignment using Wasserstein Barycenter
• A Fast Globally Linearly Convergent Algorithm for the Computation of Wasserstein Barycenters
• Estimating Barycenters of Measures in High Dimensions
• Interior-Point Methods Strike Back: Solving the Wasserstein Barycenter Problem
• Wasserstein Measure Coresets
• Projection Robust Wasserstein Barycenters
• Augmented Sliced Wasserstein Distances
• Variational Wasserstein Barycenters for Geometric Clustering
• Stochastic Saddle-Point Optimization for Wasserstein Barycenters
• Wasserstein k-means with sparse simplex projection
• Fixed Support Tree-Sliced Wasserstein Barycenter
• Learning Optimal Transport Between two Empirical Distributions with Normalizing Flows
• Approximation of Wasserstein distance with Transshipment
• Variational Wasserstein Barycenters with c-Cyclical Monotonicity
• Efficient estimates of optimal transport via low-dimensional embeddings