Revisiting Fixed Support Wasserstein Barycenter: Computational Hardness and Efficient Algorithms
share
We study the fixed-support Wasserstein barycenter problem (FS-WBP), which consists in computing the Wasserstein barycenter of m discrete probability measures supported on a finite metric space of size n. We show first that the constraint matrix arising from the linear programming (LP) representation of the FS-WBP is totally unimodular when m ≥ 3 and n = 2, but not totally unimodular when m ≥ 3 and n ≥ 3. This result answers an open problem, since it shows that the FS-WBP is not a minimum-cost flow problem and therefore cannot be solved efficiently using linear programming. Building on this negative result, we propose and analyze a simple and efficient variant of the iterative Bregman projection (IBP) algorithm, currently the most widely adopted algorithm to solve the FS-WBP. The algorithm is an accelerated IBP algorithm which achieves the complexity bound of O(mn^7/3/ε). This bound is better than that obtained for the standard IBP algorithm—O(mn^2/ε^2)—in terms of ε, and that of accelerated primal-dual gradient algorithm—O(mn^5/2/ε)—in terms of n. Empirical studies on simulated datasets demonstrate that the acceleration promised by the theory is real in practice.
READ FULL TEXT