Dr. Loganantharaj's personal home page can be reached from here.
|
Associate Professor, Computer Science.
B.Sc., Electrical and Electronic Engineering, U. of SriLanka, Peradeniya, 1977;
M.Eng., AIT, Bangkok, 1981; Ph.D., Computer Science, Colorado State U.,
Colorado, 1985.
|
|
|
Dr. Loganantharaj is interested in both the theory and the
applications of artificial intelligence. He is particularly interested in
constraint-based reasoning, constrained programming and the application of
artificial intelligence to solve some interesting problems in
BioInformatics. He is a director of a Laboratory for Intelligent Systems, which
was established by a grant from Louisiana Education Quality Enhancement
Program.
Constraint-based reasoning involves broad classes of problems. For example, when constraints are limited to time, the class of problem is called temporal reasoning. Time can be represented either implicitly or explicitly. He is interested in the explicit representation of qualitative and quantitative time. There are two competing representations of time: points and intervals. He believes both representations must be included so as to capture a problem's specification naturally. His other interests in temporal reasoning include efficient propagation algorithms, recurrent events, fuzzy temporal constraints and approximate reasoning. |
Constraint-based scheduling is another class of constraint-based reasoning where the constraints include resources, state variables (configuration), space and time. Optimally solving such scheduling problems is computationally intractable. Heuristics are used to direct the search towards a better solution. He is interested in discovering heuristics that solve the problems near optimally. He has been very successful in discovering such heuristics using evolutionary reinforcement learning techniques. Several empirical studies have revealed that many computationally intractable problems exhibit some "phase transition behavior," which is characterized as a rapid change in computational efforts over a narrow range of an order parameter. Based on such behavior, a problem instance can be categorized into one of the following three categories: under constrained, critically constrained and over constrained. A critically constrained instance is the hardest to solve.
|
He studies order parameters and the phase transition boundaries of a class of resource constrained scheduling problems. He is interested in solving BioInformatics problems using artificial intelligence techniques. He is particularly interested in discovering genes, and splice sites from a DNA sequence. In addition to having an extensive research and development experience in an academic environment, he has an extensive experience working with realistic problems in industries. During the 2000-2001 academic year, he worked at i2 Technologies, Cambridge Branch, as a senior optimization architect researching and developing state-of-the-art optimization software. Through NASA summer faculty fellowship program, he worked with several NASA research laboratories.
He has directed nine Ph. D. dissertations in the area of artificial intelligence and has published several conference and journal papers.
|
Last Modified: 19 August 2003.