Research Interests


Current projects

I am a graduate student in the Machine Learning Group at the University of Toronto with an additional interest in some aspects of Computer Vision. My PhD supervisor is Richard Zemel. We currently work with face images and are interested primarily in unsupervised and semi-supervised scenarios.

Here is my project report on Markovian Time Series Models for Language Identification for CSC2515, the class that is most to blame for my interest in Machine Learning. The report is in the NIPS paper style, which was part of the project requirement.

My Master's supervisor was Sheila McIlraith. The research for my Masters thesis involved problems arising in distributed logical reasoning. More specifically, we characterized query answering behavior in peer-to-peer networks in which peers are logical reasoners including their knowledge bases. Our prime concern was the development of inconsistency-tolerant query answering algorithms for such systems.

Projects I have previously been involved in

Multi-Agent Systems: Belief Reasoning (Fall 2003 to Fall 2004)
Undergraduate research with Thomas Ioerger at the Department of Computer Science at Texas A&M University. Belief reasoning for agents has so far mostly been addressed by modal logics which are computationally very hard to reason in. We explored a more practical approach in a conventional first-order logic back-chaining system such as Prolog.

Automated Reasoning: Semantically Guided Theorem Proving (Summer 2003)
Vacation Research Scholar, funded by the National ICT Australia (NICTA), with John Slaney of the Automated Reasoning Group (now in the Logic and Computation Group) within the Computer Sciences Laboratory (CSL) of the Research School of Information Sciences and Engineering (RSISE) at the Australian National University (ANU). One standard approach for automated reasoning in first-order logic is the given clause algorithm which can only take advantage of syntactic properties of logic clauses. John Slaney et al. have developed SCOTT, a theorem prover that uses semantic information of clauses as extracted by a constraint solver in addition to the syntactic information used by the given clause algorithm. In this project, I explored the use of a soft-constraint solver to supply semantic information by reimplementing SCOTT from its new components. I investigated, tuned, and documented the behavior of the resulting new system.

Multi-Agent Systems: Applications in Air Traffic Control (Fall 2003 to Fall 2004)
Undergraduate research with the group of John Valasek at the Flight Simulation Laboratory (FSL) of the Department of Aerospace Engineering at Texas A&M University. We simulated automated air traffic control for small, non-staffed airports under the NASA SATS program. In particular, I was involved with the multi-aircraft simulation software and the software agents "flying" these virtual aircraft.

Please click here for publications resulting from the above projects.



back to homepage