Summer School 2011
Photo by Böhringer
Analysis of relational data with factorization models
Abstract
-
Many important analysis problems can be described as prediction tasks over relational data. For example recommender systems, that are a crucial part in many online-shops, social websites, etc., try to predict the taste of users based on their behavior in the past. Whereas early research was driven mostly by heuristics, today's approaches are based on machine learning and statistical models. From a machine learning point of view, these problems are difficult as the variables are categorical of large domains, observations are typically sparse and often it is not even clear how to interpret the observation. In this talk, I will present my current research in factorization models for solving such tasks. Factorization models for several problem settings will be shown and the similarities and differences to standard machine learning approaches are discussed.