Suboptimality of Penalized Empirical Risk Minimization in Classification

dc.creatorLecué, Guillaume
dc.date2007-03-27
dc.date.accessioned2026-07-07T12:07:24Z
dc.date.available2026-07-07T12:07:24Z
dc.descriptionLet $\cF$ be a set of $M$ classification procedures with values in $[-1,1]$. Given a loss function, we want to construct a procedure which mimics at the best possible rate the best procedure in $\cF$. This fastest rate is called optimal rate of aggregation. Considering a continuous scale of loss functions with various types of convexity, we prove that optimal rates of aggregation can be either $((\log M)/n)^{1/2}$ or $(\log M)/n$. We prove that, if all the $M$ classifiers are binary, the (penalized) Empirical Risk Minimization procedures are suboptimal (even under the margin/low noise condition) when the loss function is somewhat more than convex, whereas, in that case, aggregation procedures with exponential weights achieve the optimal rate of aggregation.
dc.description15 pages
dc.identifierhttps://arxiv.org/abs/math/0703811
dc.identifierhttp://arxiv.org/abs/math/0703811
dc.identifier.urihttp://salesiana.dossiersoluciones.com/handle/123456789/208955
dc.subjectStatistics Theory
dc.subjectRisk Management
dc.subject62G05
dc.titleSuboptimality of Penalized Empirical Risk Minimization in Classification
dc.typetext

Files

Collections