Non-Parametric Modeling of Partially Ranked Data

A growing number of modeling applications involve partially ranked, rather than numeric, data. Examples include voting data, psychological studies, product recommendations in online marketing, and website and ad placement in search engines. Learning models on partial rankings of n items are often of limited practical use for large n due to computational considerations. We explore several non-parametric and conditional models for partially ranked data and derive computationally efficient procedures for large n. The derivations are largely possible through the combinatorics of the lattice of partial rankings and the algebraic structure of the symmetric group. We demonstrate the effectiveness of the proposed framework with a bias-variance analysis and a large scale experimental study involving voting data, product recommendations, and web search.

Bio
Guy Lebanon is an assistant professor at Purdue University with a joint appointment in Statistics and Electrical and Computer Engineering. His research area includes machine learning, computational data analysis, and visualization. Prof. Lebanon received the 2007 Teaching for Tomorrow Award from Purdue University and the Best Presentation Award in the 2004LTI Student Research Symposium. Prof. Lebanon received his PhD in 2005 from the Language Technologies Institute, Carnegie Mellon University anda BA and MS degrees from Technion - Israel Institute of Technology.