Show simple item record

dc.contributor.authorCaponnetto, Andrea
dc.contributor.authorRakhlin, Alexander
dc.date.accessioned2005-12-22T02:29:32Z
dc.date.available2005-12-22T02:29:32Z
dc.date.issued2005-05-17
dc.identifier.otherMIT-CSAIL-TR-2005-033
dc.identifier.otherAIM-2005-018
dc.identifier.otherCBCL-250
dc.identifier.urihttp://hdl.handle.net/1721.1/30545
dc.description.abstractWe study properties of algorithms which minimize (or almost minimize) empirical error over a Donsker class of functions. We show that the L2-diameter of the set of almost-minimizers is converging to zero in probability. Therefore, as the number of samples grows, it is becoming unlikely that adding a point (or a number of points) to the training set will result in a large jump (in L2 distance) to a new hypothesis. We also show that under some conditions the expected errors of the almost-minimizers are becoming close with a rate faster than n^{-1/2}.
dc.format.extent9 p.
dc.format.extent7033622 bytes
dc.format.extent434782 bytes
dc.format.mimetypeapplication/postscript
dc.format.mimetypeapplication/pdf
dc.language.isoen_US
dc.relation.ispartofseriesMassachusetts Institute of Technology Computer Science and Artificial Intelligence Laboratory
dc.subjectAI
dc.subjectempirical risk minimization
dc.subjectstability
dc.subjectempirical processes
dc.titleSome Properties of Empirical Risk Minimization over Donsker Classes


Files in this item

Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record