Show simple item record

dc.contributor.authorGower, Robert M.
dc.contributor.authorKovalev, Dmitry
dc.contributor.authorLieder, Felix
dc.contributor.authorRichtarik, Peter
dc.date.accessioned2019-11-27T11:17:18Z
dc.date.available2019-11-27T11:17:18Z
dc.date.issued2019-05-26
dc.identifier.urihttp://hdl.handle.net/10754/660275
dc.description.abstractWe develop a randomized Newton method capable of solving learning problems with huge dimensional feature spaces, which is a common setting in applications such as medical imaging, genomics and seismology. Our method leverages randomized sketching in a new way, by finding the Newton direction constrained to the space spanned by a random sketch. We develop a simple global linear convergence theory that holds for practically all sketching techniques, which gives the practitioners the freedom to design custom sketching approaches suitable for particular applications. We perform numerical experiments which demonstrate the efficiency of our method as compared to accelerated gradient descent and the full Newton method. Our method can be seen as a refinement and randomized extension of the results of Karimireddy, Stich, and Jaggi (2019).
dc.publisherarXiv
dc.relation.urlhttps://arxiv.org/pdf/1905.10874
dc.rightsArchived with thanks to arXiv
dc.titleRSN: Randomized Subspace Newton
dc.typePreprint
dc.contributor.departmentKAUST, Saudi Arabia
dc.contributor.departmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
dc.contributor.departmentComputer Science Program
dc.eprint.versionPre-print
dc.contributor.institutionLTCI, Telecom Paristech, IPP, France
dc.contributor.institutionHeinrich-Heine-Universitat Dusseldorf, Germany
dc.contributor.institutionMIPT, Russia
dc.identifier.arxivid1905.10874
kaust.personKovalev, Dmitry
kaust.personRichtarik, Peter
refterms.dateFOA2019-11-27T11:17:45Z


Files in this item

Thumbnail
Name:
Preprintfile1.pdf
Size:
889.3Kb
Format:
PDF
Description:
Pre-print

This item appears in the following Collection(s)

Show simple item record