People
Faculty Members
Ronald J. Williams |
 |
Professor Williams conducts research in the field of machine learning, a subspecialty of artificial intelligence. His current focus is the development and analysis of algorithms for autonomous agents facing delayed-reinforcement learning tasks. Such tasks involve choosing sequences of actions to accomplish desired objectives when facing varying amounts of initial uncertainty about the effects of an agent's actions. Developing strategies for performing these tasks effectively incorporates elements of both learning and planning. Many of the algorithms Professor Williams studies are variants of existing approaches used in operations research and optimal control theory. These also incorporate such features as function approximators, including artificial neural networks. In his current research, Professor Williams considers how agents can learn as rapidly as possible from limited interaction with their environment.
Earlier, Professor Williams researched approaches to training artificial neural networks to perform varied tasks. He was a codeveloper of the widely used back-propagation algorithm used to train feed-forward nets and also investigated techniques to train recurrent networks to simulate dynamical systems.
Professor Williams has served as associate editor of Adaptive Behavior and IEEE Transactions in Natural Computation, and is on the editorial boards of both Machine Learning and Neural Computation.
Career Publication Highlights
Williams, Ronald J. 1992. Simple statistical gradient-following algorithms for connectionist reinforcement learning. Machine Learning 8:229-256.
Williams, Ronald J., David E. Rumelhart, and Geoffrey E. Hinton. 1986. Learning internal representations by error propagation. In Parallel distributed processing: Explorations in the microstructure of cognition, ed. David E. Rumelhart and J. A. McClelland. Vol. 1. Cambridge: MIT Press/Bradford Books.
Williams, Ronald J., and David Zipser. 1995. Gradient-based learning algorithms for recurrent networks and their computational complexity. In Back propagation: Theory, architectures, and applications, ed. Yves Chauvin and David E. Rumelhart. Hillsdale, N.J.: Erlbaum