Strong Asymptotic Assertions for Discrete MDL in Regression and Classification

Simple item page
dc.creatorPoland, Jan
dc.creatorHutter, Marcus
dc.date2005-02-15
dc.date.accessioned2026-07-07T08:18:16Z
dc.date.available2026-07-07T08:18:16Z
dc.descriptionWe study the properties of the MDL (or maximum penalized complexity) estimator for Regression and Classification, where the underlying model class is countable. We show in particular a finite bound on the Hellinger losses under the only assumption that there is a "true" model contained in the class. This implies almost sure convergence of the predictive distribution to the true one at a fast rate. It corresponds to Solomonoff's central theorem of universal induction, however with a bound that is exponentially larger.
dc.description6 two-column pages
dc.identifierhttps://arxiv.org/abs/math/0502315
dc.identifierhttp://arxiv.org/abs/math/0502315
dc.identifierProc. 14th Dutch-Belgium Conf. on Machine Learning (Benelearn 2005) 67-72
dc.identifier.urihttp://salesiana.dossiersoluciones.com/handle/123456789/134376
dc.subjectStatistics Theory
dc.subjectArtificial Intelligence
dc.subjectInformation Theory
dc.subjectMachine Learning
dc.subjectProbability
dc.titleStrong Asymptotic Assertions for Discrete MDL in Regression and Classification
dc.typetext

Files

Collections

arXiv