Clare Bates Congdon
Location: (Portland, ME)
Personal Research Web Page: http://www.cs.usm.maine.edu/~congdon/
Keywords: Evolutionary computation and machine learning applied to bioinformatics and computational biology, including DNA sequence analysis and phylogenetics, and applied to intelligent agents, including robotics and interactive dynamic games.
Posted on: Thursday, May 5th, 2011
Broad Research Area: AI / Machine Learning / Robotics / Vision, Other, Scientific/Medical Informatics
Research Interests:
The Bioinformatics and Intelligent Systems Lab at USM focuses on evolutionary computation and other artificial intelligence approaches as applied primarily to bioinformatics and intelligent agents.
Bioinformatics projects include:
GAMI, a genetic algorithms approach to DNA motif inference: In this project we identify patterns in noncoding DNA that have been conserved across evolutionary time; such elements are good candidates for affecting the function of genes and will be studied at the bench by our biological collaborators. This project is motivated by a study of CFTR, the gene for cystic fibrosis and has expanded to other environmental-response genes; among GAMI’s strengths is that it is designed to work with long sequences (e.g., 100k) and many of them (e.g., 100 or more) There are many subprojects here; one particularly important horizon is that functional elements in noncoding DNA tend to appear in modules (not in isolation), and we will design a new system to infer these.
Gaphyl, a genetic algorithms approach to phylogenetics, the inference of trees representing the evolutionary relationships among species or strains. We are just starting a swine flu project, which will also mean designing a new computational approach designed to work specifically with viruses.
My bioinformatics work is collaborative with researchers at the Mount Desert Island Biological Lab, Dartmouth College, the University of Maine (Orono), and the University of Illinois.
Intelligent agents:
Our recent intelligent agents work has focused on game-playing agents. We won first place in the 2008 Ms. PacMan competition at the World Congress on Computational Intelligence (Hong Kong), the 2009 Unreal Tournament competition at the Congress on Evolutionary Computation (Trondheim, Norway), and in both the Learning Track and Gameplay Track of the 2010 Computational Intelligence and Games Conference (Copenhagen, Denmark). Designing agents for interactive dynamic games is a very similar research problem to real-time robotics, where inputs must quickly be processed to determine appropriate outputs, and split-second decisions must be made. Our approach is adaptive rule-based systems.
I am also working with Jim Wilson and others at the University of Maine on a Complex Adaptive Systems project, using learning classifier systems to model cooperation and communication in the lobster, sea urchin, and groundfish fisheries.
Other projects include:
Lobster CyberCatch, an educational science game designed to teach middle-school students about math and science concepts in general and lobsters (and lobstering) more specifically.
VIEWER, a project centered at the University of Maine (Orono) to develop touch-screen visualization walls for remote conferencing for scientific collaborations, including informal “water cooler” conversations.
Contact Information:
Clare Bates Congdon Associate Professor Department of Computer Science University of Southern Maine P.O. Box 9300, Portland, ME 04104-9300 congdon@usm.maine.edu http://www.cs.usm.maine.edu/~congdon/
