Robustness in Compositional Coevolution
Dr. R. Paul Wiegand
Friday, September 28th, 2007
1:00 PM ~ 2:00 PM
Harris Engineering Center 101
Abstract:
Though recent analysis of traditional compositional coevolutionary algorithms (CCEAs) casts doubt on their suitability for static optimization tasks, empirical evidence suggests that the algorithms perform quite well in multiagent learning settings. This is due in part because many CCEAs may be quite suitable to finding behaviors for team members that result in good (though not necessarily optimal) performance but which are also robust to changes in other team members. Given this, there are two main goals of this talk. First, I will describe a general framework for clearly defining robustness, offering a specific definition for our studies. Second, I will examine the hypothesis that CCEAs exploit this robustness property during their search. I use an existing theoretical model to gain intuition about the kind of problem properties that attract populations in the system, then provide a simple empirical study justifying this intuition in a practical setting. Moreover, I will outline early steps toward a theoretical foundation for this work. The goal of these efforts is to develop a constructive view of CCEAs as optimizers of robustness.
Bio:
R. Paul Wiegand received his Ph.D. from George Mason University in 2004, on the topic of An Analysis of Cooperative Coevolutionary Algorithms. He currently holds a research faculty position at the Institute for Simulation & Training at UCF, and has conducted research at the Navy Center for Applied Research in Artificial Intelligence at the Naval Research Laboratory, the Adaptive Systems and Evolutionary Computation laboratories at George Mason University, and the Algorithm Efficiency and Complexity group at Dortmund Universit=E4t. His research interests primarily focus on methods of natural computation, theory of coadaptive and coevolutionary computation, and application of coadaptive methods for multiagent learning. More generally, Dr. Wiegand is interested in studying methods for designing and applying effective learning algorithms and representations for generating and modeling robust heterogeneous, multiagent team behaviors.
Dr. Wiegand is in the early stages of developing the Natural Computation & Coadaptive Systems Laboratory at IST. The focus of this group will be to study computational mechanisms modeled after natural phenomena (e.g., evolutionary computation, physics- or biologically-based swarm systems, ant colony optimization, etc.), as well as complex adaptive systems that involve multiple, interacting entities (e.g., coevolution, multiagent reinforcement learning, swarm behaviors, etc.). The primary domains of interest will be multiagent simulation, and there will be strong analytical component to the research conducted by the lab. Those interested in being involved, particularly Computer Science and Modeling & Simulation students at UCF, are encouraged to contact Dr. Wiegand.
