Loading…

Interpolating between Optimal Transport and MMD using Sinkhorn Divergences

Comparing probability distributions is a fundamental problem in data sciences. Simple norms and divergences such as the total variation and the relative entropy only compare densities in a point-wise manner and fail to capture the geometric nature of the problem. In sharp contrast, Maximum Mean Disc...

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2018-10
Main Authors: Feydy, Jean, Thibault Séjourné, Vialard, François-Xavier, Amari, Shun-ichi, Trouvé, Alain, Peyré, Gabriel
Format: Article
Language:English
Subjects:
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by
cites
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Feydy, Jean
Thibault Séjourné
Vialard, François-Xavier
Amari, Shun-ichi
Trouvé, Alain
Peyré, Gabriel
description Comparing probability distributions is a fundamental problem in data sciences. Simple norms and divergences such as the total variation and the relative entropy only compare densities in a point-wise manner and fail to capture the geometric nature of the problem. In sharp contrast, Maximum Mean Discrepancies (MMD) and Optimal Transport distances (OT) are two classes of distances between measures that take into account the geometry of the underlying space and metrize the convergence in law. This paper studies the Sinkhorn divergences, a family of geometric divergences that interpolates between MMD and OT. Relying on a new notion of geometric entropy, we provide theoretical guarantees for these divergences: positivity, convexity and metrization of the convergence in law. On the practical side, we detail a numerical scheme that enables the large scale application of these divergences for machine learning: on the GPU, gradients of the Sinkhorn loss can be computed for batches of a million samples.
format article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2124004731</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2124004731</sourcerecordid><originalsourceid>FETCH-proquest_journals_21240047313</originalsourceid><addsrcrecordid>eNqNykEKwjAQQNEgCBbtHQKuC2nSWvdWUaG40H2JOtbUOomTVK-vggdw9Rf_DVgklUqTeSbliMXet0IIOStknquIbTcYgJztdDDY8COEFwDynQvmrjt-II3eWQpc45lXVcl7_3V7g7erJeSleQI1gCfwEza86M5D_OuYTVfLw2KdOLKPHnyoW9sTflYtU5kJkRUqVf-pN-8qPN4</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2124004731</pqid></control><display><type>article</type><title>Interpolating between Optimal Transport and MMD using Sinkhorn Divergences</title><source>Publicly Available Content Database</source><creator>Feydy, Jean ; Thibault Séjourné ; Vialard, François-Xavier ; Amari, Shun-ichi ; Trouvé, Alain ; Peyré, Gabriel</creator><creatorcontrib>Feydy, Jean ; Thibault Séjourné ; Vialard, François-Xavier ; Amari, Shun-ichi ; Trouvé, Alain ; Peyré, Gabriel</creatorcontrib><description>Comparing probability distributions is a fundamental problem in data sciences. Simple norms and divergences such as the total variation and the relative entropy only compare densities in a point-wise manner and fail to capture the geometric nature of the problem. In sharp contrast, Maximum Mean Discrepancies (MMD) and Optimal Transport distances (OT) are two classes of distances between measures that take into account the geometry of the underlying space and metrize the convergence in law. This paper studies the Sinkhorn divergences, a family of geometric divergences that interpolates between MMD and OT. Relying on a new notion of geometric entropy, we provide theoretical guarantees for these divergences: positivity, convexity and metrization of the convergence in law. On the practical side, we detail a numerical scheme that enables the large scale application of these divergences for machine learning: on the GPU, gradients of the Sinkhorn loss can be computed for batches of a million samples.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Convergence ; Convexity ; Entropy ; Machine learning ; Norms</subject><ispartof>arXiv.org, 2018-10</ispartof><rights>2018. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://www.proquest.com/docview/2124004731?pq-origsite=primo$$EHTML$$P50$$Gproquest$$Hfree_for_read</linktohtml><link.rule.ids>780,784,25753,37012,44590</link.rule.ids></links><search><creatorcontrib>Feydy, Jean</creatorcontrib><creatorcontrib>Thibault Séjourné</creatorcontrib><creatorcontrib>Vialard, François-Xavier</creatorcontrib><creatorcontrib>Amari, Shun-ichi</creatorcontrib><creatorcontrib>Trouvé, Alain</creatorcontrib><creatorcontrib>Peyré, Gabriel</creatorcontrib><title>Interpolating between Optimal Transport and MMD using Sinkhorn Divergences</title><title>arXiv.org</title><description>Comparing probability distributions is a fundamental problem in data sciences. Simple norms and divergences such as the total variation and the relative entropy only compare densities in a point-wise manner and fail to capture the geometric nature of the problem. In sharp contrast, Maximum Mean Discrepancies (MMD) and Optimal Transport distances (OT) are two classes of distances between measures that take into account the geometry of the underlying space and metrize the convergence in law. This paper studies the Sinkhorn divergences, a family of geometric divergences that interpolates between MMD and OT. Relying on a new notion of geometric entropy, we provide theoretical guarantees for these divergences: positivity, convexity and metrization of the convergence in law. On the practical side, we detail a numerical scheme that enables the large scale application of these divergences for machine learning: on the GPU, gradients of the Sinkhorn loss can be computed for batches of a million samples.</description><subject>Convergence</subject><subject>Convexity</subject><subject>Entropy</subject><subject>Machine learning</subject><subject>Norms</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2018</creationdate><recordtype>article</recordtype><sourceid>PIMPY</sourceid><recordid>eNqNykEKwjAQQNEgCBbtHQKuC2nSWvdWUaG40H2JOtbUOomTVK-vggdw9Rf_DVgklUqTeSbliMXet0IIOStknquIbTcYgJztdDDY8COEFwDynQvmrjt-II3eWQpc45lXVcl7_3V7g7erJeSleQI1gCfwEza86M5D_OuYTVfLw2KdOLKPHnyoW9sTflYtU5kJkRUqVf-pN-8qPN4</recordid><startdate>20181018</startdate><enddate>20181018</enddate><creator>Feydy, Jean</creator><creator>Thibault Séjourné</creator><creator>Vialard, François-Xavier</creator><creator>Amari, Shun-ichi</creator><creator>Trouvé, Alain</creator><creator>Peyré, Gabriel</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20181018</creationdate><title>Interpolating between Optimal Transport and MMD using Sinkhorn Divergences</title><author>Feydy, Jean ; Thibault Séjourné ; Vialard, François-Xavier ; Amari, Shun-ichi ; Trouvé, Alain ; Peyré, Gabriel</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_21240047313</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2018</creationdate><topic>Convergence</topic><topic>Convexity</topic><topic>Entropy</topic><topic>Machine learning</topic><topic>Norms</topic><toplevel>online_resources</toplevel><creatorcontrib>Feydy, Jean</creatorcontrib><creatorcontrib>Thibault Séjourné</creatorcontrib><creatorcontrib>Vialard, François-Xavier</creatorcontrib><creatorcontrib>Amari, Shun-ichi</creatorcontrib><creatorcontrib>Trouvé, Alain</creatorcontrib><creatorcontrib>Peyré, Gabriel</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni)</collection><collection>ProQuest Central</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Feydy, Jean</au><au>Thibault Séjourné</au><au>Vialard, François-Xavier</au><au>Amari, Shun-ichi</au><au>Trouvé, Alain</au><au>Peyré, Gabriel</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Interpolating between Optimal Transport and MMD using Sinkhorn Divergences</atitle><jtitle>arXiv.org</jtitle><date>2018-10-18</date><risdate>2018</risdate><eissn>2331-8422</eissn><abstract>Comparing probability distributions is a fundamental problem in data sciences. Simple norms and divergences such as the total variation and the relative entropy only compare densities in a point-wise manner and fail to capture the geometric nature of the problem. In sharp contrast, Maximum Mean Discrepancies (MMD) and Optimal Transport distances (OT) are two classes of distances between measures that take into account the geometry of the underlying space and metrize the convergence in law. This paper studies the Sinkhorn divergences, a family of geometric divergences that interpolates between MMD and OT. Relying on a new notion of geometric entropy, we provide theoretical guarantees for these divergences: positivity, convexity and metrization of the convergence in law. On the practical side, we detail a numerical scheme that enables the large scale application of these divergences for machine learning: on the GPU, gradients of the Sinkhorn loss can be computed for batches of a million samples.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2018-10
issn 2331-8422
language eng
recordid cdi_proquest_journals_2124004731
source Publicly Available Content Database
subjects Convergence
Convexity
Entropy
Machine learning
Norms
title Interpolating between Optimal Transport and MMD using Sinkhorn Divergences
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T17%3A18%3A44IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Interpolating%20between%20Optimal%20Transport%20and%20MMD%20using%20Sinkhorn%20Divergences&rft.jtitle=arXiv.org&rft.au=Feydy,%20Jean&rft.date=2018-10-18&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2124004731%3C/proquest%3E%3Cgrp_id%3Ecdi_FETCH-proquest_journals_21240047313%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=2124004731&rft_id=info:pmid/&rfr_iscdi=true