Marginal Constrained Latent Variable Modeling for Ranking and Collaborative Filtering Draft

  • Published 2007

Abstract

Latent or hidden variable based statistical methods constitute a set of valuable techniques for modeling co-occurrence data, as they can effectively trade-off the simplicity of assuming independence among the features with the computational intractability of modeling a full joint distribution. Such models however are primarily “explanatory” rather than… (More)

Topics

  • Presentations referencing similar topics