Lars Otten Lars Otten

Position:  Graduate student
Email: email address (GPG key)
Office: 4099 Bren Hall
Address:  Department of Computer Science
University of California, Irvine
Irvine, CA  92697-3435

I am a graduate student in Prof. Rina Dechter's group, working towards a Ph.D. in the School of Information and Computer Science at UC Irvine since the fall of 2006.

 

Research

My area of concentration is artificial intelligence in general and the graphical model framework in particular, where I am trying to exploit structure in problem solving. The reasoning tasks I deal with comprise constraint satisfaction and optimization problems as well as queries over Bayesian and Markov networks (e.g., P(e), MPE, belief updating).

I have previously investigated different complexity measures for reasoning problems represented as graphical models. Specifically, I worked on instance-based complexity bounds in the presence of determinism, which can be used to guide parameter selection for the underlying algorithm.

More recently, my focus has shifted a bit, intersecting AI with parallel and distributed computing: My current research circles around parallelization strategies for established state-of-the-art algorithms in the context of graphical model reasoning. At this point, the focus lies on computational grid environments, but other systems are under consideration for future research as well.

Applications are manifold, but one driving factor in our experimental evaluation are haplotyping and linkage problems from human genetic analysis. These can be modeled as Bayesian networks and are thus susceptible to our highly advanced methods like AND/OR graph search. With inherently exponential complexity, however, and evergrowing sets of available data, these kinds of problems make it very natural to exploit vast computational resources in parallel.

I've implemented a first prototype system on a small local computational grid and made it accessible through a web site interface. It uses our algorithms for finding the most likely haplotypes in general pedigrees:

At this point the system is not open to the public, please contact me directly if you are interested.

Our research group recently launched a website where we make available implementations of the algorithms we develop: http://graphmod.ics.uci.edu/.

I have been involved in the UAI'08 Probabilistic Inference Evaluation.

 

Teaching

I am not involved in any courses this quarter.

In Spring 2009 I was the teaching assistant for CS 174 "Bioinformatics". I was a reader for ICS 11 / Econ 11 "The Internet and Public Policy" in Winter 2009 and teaching assistant for CS 175 "Project in AI" in Fall 2008.

 

Personal

Before coming to UCI I was a student at RWTH Aachen University in Aachen, Germany and Chalmers University of Technology in Gothenburg, Sweden.

The best way to get in touch with me is via email at email address. For encrypted email, you can obtain the respective GnuPG key from a key server (here for instance), its ID is 0x830EB280 with the following fingerprint:
869A 685A 0951 E946 92C0  8C7F A9A6 C83A 830E B280.

Finally, if you are curious, you can visit my (non-academic) personal webpage.

 

Under Review

Refereed Conference Publications

Workshops