Fast and Memory-Efficient Algorithms for High-Order Tucker Decomposition

Cited 2 time in webofscience Cited 0 time in scopus
  • Hit : 219
  • Download : 0
DC FieldValueLanguage
dc.contributor.authorZhang, Jiyuanko
dc.contributor.authorOh, Jinohko
dc.contributor.authorShin, Kijungko
dc.contributor.authorPapalexakis, Evangelos E.ko
dc.contributor.authorFaloutsos, Christosko
dc.contributor.authorYu, Hwanjoko
dc.date.accessioned2020-10-20T01:55:32Z-
dc.date.available2020-10-20T01:55:32Z-
dc.date.created2020-01-23-
dc.date.created2020-01-23-
dc.date.created2020-01-23-
dc.date.created2020-01-23-
dc.date.issued2020-07-
dc.identifier.citationKNOWLEDGE AND INFORMATION SYSTEMS, v.62, no.7, pp.2765 - 2794-
dc.identifier.issn0219-1377-
dc.identifier.urihttp://hdl.handle.net/10203/276684-
dc.description.abstractMulti-aspect data appear frequently in web-related applications. For example, product reviews are quadruplets of the form (user, product, keyword, timestamp), and search-engine logs are quadruplets of the form (user, keyword, location, timestamp). How can we analyze such web-scale multi-aspect data on an off-the-shelf workstation with a limited amount of memory? Tucker decomposition has been used widely for discovering patterns in such multi-aspect data, which are naturally expressed as large but sparse tensors. However, existing Tucker decomposition algorithms have limited scalability, failing to decompose large-scale high-order (= 4) tensors, since they explicitly materialize intermediate data, whose size grows exponentially with the order. To address this problem, which we call "Materialization Bottleneck," we propose S- HOT, a scalable algorithm for high-order Tucker decomposition. S- HOT minimizes materialized intermediate data by using an on-the-fly computation, and it is optimized for disk-resident tensors that are too large to fit in memory. We theoretically analyze the amount of memory and the number of data scans required by S- HOT. Moreover, we empirically showthat S- HOT handles tensors with higher order, dimensionality, and rank than baselines. For example, S- HOT successfully decomposes a real-world tensor from the Microsoft Academic Graph on an off-the-shelf workstation, while all baselines fail. Especially, in terms of dimensionality, S- HOT decomposes 1000xlarger tensors than baselines.-
dc.languageEnglish-
dc.publisherSPRINGER LONDON LTD-
dc.titleFast and Memory-Efficient Algorithms for High-Order Tucker Decomposition-
dc.typeArticle-
dc.identifier.wosid000515603900001-
dc.identifier.scopusid2-s2.0-85078491286-
dc.type.rimsART-
dc.citation.volume62-
dc.citation.issue7-
dc.citation.beginningpage2765-
dc.citation.endingpage2794-
dc.citation.publicationnameKNOWLEDGE AND INFORMATION SYSTEMS-
dc.identifier.doi10.1007/s10115-019-01435-1-
dc.contributor.localauthorShin, Kijung-
dc.contributor.nonIdAuthorZhang, Jiyuan-
dc.contributor.nonIdAuthorOh, Jinoh-
dc.contributor.nonIdAuthorPapalexakis, Evangelos E.-
dc.contributor.nonIdAuthorFaloutsos, Christos-
dc.contributor.nonIdAuthorYu, Hwanjo-
dc.description.isOpenAccessN-
dc.type.journalArticleArticle-
dc.subject.keywordAuthorTensor-
dc.subject.keywordAuthorHigh-order tensor-
dc.subject.keywordAuthorTensor decomposition-
dc.subject.keywordAuthorTucker decomposition-
dc.subject.keywordAuthorOut-of-core algorithm-
Appears in Collection
AI-Journal Papers(저널논문)
Files in This Item
There are no files associated with this item.
This item is cited by other documents in WoS
⊙ Detail Information in WoSⓡ Click to see webofscience_button
⊙ Cited 2 items in WoS Click to see citing articles in records_button

qr_code

  • mendeley

    citeulike


rss_1.0 rss_2.0 atom_1.0