Communications Seminar

Title:  Active Learning for Online Bayesian Matrix Factorization

Speaker:  Dr. Jorge G. Silva

Electrical & Computer Eng.

Duke University

Date:  Monday, September 17, 2012

Time:  4:00 p.m.

Location:  141 Coordinated Science Lab

Abstract:  The problem of large-scale online matrix completion is addressed via a Bayesian approach. The proposed method learns a factor analysis (FA) model for large matrices, based on a small number of observed matrix elements, and leverages the statistical model to actively select which new matrix entries/observations would be most informative if they could be acquired, to improve the model; the model inference and active learning are performed in an online setting. In the context of online learning, a greedy, fast and provably near-optimal algorithm is employed to sequentially maximize the mutual information between past and future observations, taking advantage of submodularity properties. Additionally, a simpler procedure, which directly uses the posterior parameters learned by the Bayesian approach, is shown to achieve slightly lower estimation quality, with far less computational effort. Inference is performed using a computationally efficient online variational Bayes (VB) procedure. Competitive results are obtained in a very large collaborative filtering problem, namely the Yahoo! Music ratings dataset.

Biography:  Jorge Silva received his EE, MSc and PhD in Electrical and Computer Engineering from Instituto Superior Técnico  (IST), Lisbon, Portugal, in 1993, 1999 and 2007, respectively. He was a researcher at Instituto de Engenharia de Sistemas e Computadores (INESC) in 1993--1996, under a PRAXIS XXI award and at the Instituto de Sistemas e Robótica (ISR), Lisbon, in 2003--2007. He was a Teaching Assistant and later Adjunct Professor at Instituto Superior de Engenharia de Lisboa (ISEL) in 1996--2007. In the same period, he did consulting and R&D work for major Portuguese utility and transportation companies. He is currently a Senior Research Scientist at Duke University, where he is working on statistical models for very high-dimensional data. His research interests include manifold learning, kernel methods, nonlinear prediction and filtering and computer vision.