I'm in my sixth year as a PhD student, working in the general area of machine learning.

My current work focuses on making fundamental methods fast/scalable for massive datasets using Monte Carlo principles. For instance, we have fast Monte Carlo versions of the nonparametric kernel methods (KDE, kernel regression, etc.), and are working on fast Monte Carlo linear algebraic operations such as the SVD (which would apply to PCA, etc.).

Other work includes prediction in time series and dynamical systems, the Netflix prize competition (collaborative filtering), and hidden state discovery.

My advisor is Charles Isbell, and I also do work with Alex Gray.

Some papers of interest

Ultrafast Monte Carlo for Kernel Estimators and Generalized Statistical Summations
(NIPS 2007)

Fast Nonparametric Conditional Density Estimation
(UAI 2007)
Gives a method for fast bandwidth selection in kernel conditional density estimation, which was previously unscalable. Speedups as high as 10^6 are observed. The key idea is a dual-tree recursion with bootstrapped (sampled) rather than deterministic error bounding.

Looping Suffix Tree-Based Inference of Partially Observable Hidden State
(ICML 2006, Distinguished Student Paper Award)
Gives a history-based solution to the problem of inferring a complete predictive state model in deterministic POMDPs. We introduce the looping suffix tree representation, which allows us to give a unified view of other hidden state methods (including schema learning) in the context of DPOMDPs.

Schema Learning: Experience-Based Construction of Predictive Action Models
(NIPS 2004)
A constructivist approach that incrementally builds up predictors of action effects.

Audio Ad Filter
Combines an SVM with an HMM to recognize radio ads with the intent of filtering them out.

Contact Info
mph@cc.gatech.edu
fax. 404-894-9442


© 2002-2007 Michael Holmes. All Rights Reserved.