Show simple item record

dc.contributor.authorMishchenko, Konstantin
dc.date.accessioned2019-12-24T08:00:34Z
dc.date.available2019-12-24T08:00:34Z
dc.date.issued2019-09-16
dc.identifier.urihttp://hdl.handle.net/10754/660769
dc.description.abstractWe present a new perspective on the celebrated Sinkhorn algorithm by showing that is a special case of incremental/stochastic mirror descent. In order to see this, one should simply plug Kullback-Leibler divergence in both mirror map and the objective function. Since the problem has unbounded domain, the objective function is neither smooth nor it has bounded gradients. However, one can still approach the problem using the notion of relative smoothness, obtaining that the stochastic objective is 1-relative smooth. The discovered equivalence allows us to propose 1) new methods for optimal transport, 2) an extension of Sinkhorn algorithm beyond two constraints.
dc.publisherarXiv
dc.relation.urlhttps://arxiv.org/pdf/1909.06918
dc.rightsArchived with thanks to arXiv
dc.titleSinkhorn Algorithm as a Special Case of Stochastic Mirror Descent
dc.typePreprint
dc.contributor.departmentComputer Science Program
dc.eprint.versionPre-print
dc.identifier.arxivid1909.06918
kaust.personMishchenko, Konstantin
refterms.dateFOA2019-12-24T08:01:05Z


Files in this item

Thumbnail
Name:
Preprintfile1.pdf
Size:
135.2Kb
Format:
PDF
Description:
Pre-print

This item appears in the following Collection(s)

Show simple item record