We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

math.OC

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Optimization and Control

Title: Stochastic Approximation versus Sample Average Approximation for population Wasserstein barycenters

Abstract: In the machine learning and optimization community, there are two main approaches for the convex risk minimization problem, namely, the Stochastic Approximation (SA) and the Sample Average Approximation (SAA). In terms of oracle complexity (required number of stochastic gradient evaluations), both approaches are considered equivalent on average (up to a logarithmic factor). The total complexity depends on the specific problem, however, starting from work \cite{nemirovski2009robust} it was generally accepted that the SA is better than the SAA. % Nevertheless, in case of large-scale problems SA may run out of memory as storing all data on one machine and organizing online access to it can be impossible without communications with other machines. SAA in contradistinction to SA allows parallel/distributed calculations. We show that for the Wasserstein barycenter problem this superiority can be inverted. We provide a detailed comparison by stating the complexity bounds for the SA and the SAA implementations calculating barycenters defined with respect to optimal transport distances and entropy-regularized optimal transport distances. As a byproduct, we also construct confidence intervals for the barycenter defined with respect to entropy-regularized optimal transport distances in the $\ell_2$-norm. The preliminary results are derived for a general convex optimization problem given by the expectation in order to have other applications besides the Wasserstein barycenter problem.
Comments: 33 pages
Subjects: Optimization and Control (math.OC); Machine Learning (cs.LG); Machine Learning (stat.ML)
Cite as: arXiv:2001.07697 [math.OC]
  (or arXiv:2001.07697v9 [math.OC] for this version)

Submission history

From: Darina Dvinskikh [view email]
[v1] Tue, 21 Jan 2020 18:54:39 GMT (30kb)
[v2] Thu, 30 Jan 2020 14:56:01 GMT (30kb)
[v3] Mon, 18 May 2020 21:29:32 GMT (124kb,D)
[v4] Mon, 8 Jun 2020 10:02:53 GMT (816kb,D)
[v5] Wed, 16 Sep 2020 13:33:01 GMT (848kb,D)
[v6] Wed, 21 Oct 2020 13:02:59 GMT (828kb,D)
[v7] Thu, 26 Nov 2020 16:09:07 GMT (840kb,D)
[v8] Tue, 1 Dec 2020 14:34:25 GMT (840kb,D)
[v9] Mon, 25 Oct 2021 14:16:23 GMT (3638kb,D)

Link back to: arXiv, form interface, contact.