Research Interests

Computational Biology, Biomedical Informatics: Data Mining and Information Retrieval
Machine Learning
Databases -- Similarity Queries


Computational Biology, Biomedical Informatics: Data Mining and Information Retrieval

My work concentrates on applying machine learning techniques to biomedical data. Specific current projects include informative/functional SNP selection for disease-gene association, studying the combination of environmental and genetic factors involved in Autism (these belong in what is currently referred to as Translational Biomedicine), protein sub-cellular localization (sorting), and biomedical literature mining - finding documents and facts of interest in the PubMed database (and/or other text sources) using a wide variety of tools and approaches. A lot of the work is done within a probabilistic Bayesian framework, where we formulate the problems of finding relevant and significant information as a search for parameters of a specifically-defined probabilistic model.
I also work on other interesting problems in genomic sequence analysis. A lot of the work is done collaboratively, with researchers in Canada, North America, and Europe.
: Some related publications: (See the publications page for a complete list of relevant publications)


Machine Learning (Earlier work)

Machine Learning for robotics and for other dynamical systems. My PhD work was about learning maps, represented as collections of hidden Markov models, (such collections are aka POMDP models), for robot navigation purposes. Like in other forms of learning models for dynamical systems, the input is a sequence of data, and the learning task is to fit a model to the data. In the robot navigation case, the input is a sequence of observations gathered by the robot, and the output is a POMDP representing the environment traversed by the robot. This work was done with my PhD advisor, Prof. Leslie Kaelbling.

Related publications:


Similarity Queries (Even earlier...)

Looking in a dataset for patterns that "sort-of-look-like-this"... I still find this problem interesting and relevant. Related publications:

Last modified: July 4, 2008