Multiplicative updates for nonnegative quadratic programming in support vector machines

Simple item page
dc.contributor.authorSha, Fei
dc.contributor.authorSaul, Lawrence K
dc.contributor.authorLee, Daniel D
dc.date2023-05-16T21:42:19.000
dc.date.accessioned2023-05-22T12:47:19Z
dc.date.available2023-05-22T12:47:19Z
dc.date.issued2002-12-10
dc.date.submitted2004-11-11T12:48:44-08:00
dc.description.abstractWe derive multiplicative updates for solving the nonnegative quadratic programming problem in support vector machines (SVMs). The updates have a simple closed form, and we prove that they converge monotonically to the solution of the maximum margin hyperplane. The updates optimize the traditionally proposed objective function for SVMs. They do not involve any heuristics such as choosing a learning rate or deciding which variables to update at each iteration. They can be used to adjust all the quadratic programming variables in parallel with a guarantee of improvement at each iteration. We analyze the asymptotic convergence of the updates and show that the coefficients of non-support vectors decay geometrically to zero at a rate that depends on their margins. In practice, the updates converge very rapidly to good classifiers.
dc.description.commentsAdvances in Neural Information Processing Systems 15 (NIPS 2002), pages 1065-1072. Publisher URL: http://books.nips.cc/nips15.html">http://books.nips.cc/nips15.html
dc.identifier.urihttps://repository.upenn.edu/handle/20.500.14332/6329
dc.legacy.articleid1036
dc.legacy.fulltexturlhttps://repository.upenn.edu/cgi/viewcontent.cgi?article=1036&context=cis_papers&unstamped=1
dc.source.issue29
dc.source.journalDepartmental Papers (CIS)
dc.source.peerreviewedtrue
dc.source.statuspublished
dc.titleMultiplicative updates for nonnegative quadratic programming in support vector machines
dc.typePresentation
digcom.contributor.authorisAuthorOfPublication|email:feisha@seas.upenn.edu|institution:University of Pennsylvania|Sha, Fei
digcom.contributor.authorisAuthorOfPublication|email:lsaul@cis.upenn.edu|institution:University of Pennsylvania|Saul, Lawrence K
digcom.contributor.authorisAuthorOfPublication|email:ddlee@seas.upenn.edu|institution:University of Pennsylvania|Lee, Daniel D
digcom.identifiercis_papers/29
digcom.identifier.contextkey32466
digcom.identifier.submissionpathcis_papers/29
digcom.typeconference
dspace.entity.typePublication
relation.isAuthorOfPublication174dd665-ef23-479c-93b5-160050a9a576
relation.isAuthorOfPublicationa46f7bc2-dc15-4b9e-a854-f0055d4d4a2f
relation.isAuthorOfPublication044c3dcb-cd50-41da-8642-3534e2c77660
relation.isAuthorOfPublication.latestForDiscovery174dd665-ef23-479c-93b5-160050a9a576
upenn.schoolDepartmentCenterDepartmental Papers (CIS)
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
multiplicative.pdf
Size:
90.04 KB
Format:
Adobe Portable Document Format
Download
Collection
Presentations