Date: Thu, 21 Nov 1996 23:24:51 GMT
Server: NCSA/1.5.2
Last-modified: Tue, 03 Sep 1996 18:37:57 GMT
Content-type: text/html
Content-length: 2136
Michael Duff
Michael Duff
Much of my work involves bringing techniques of classical applied
probability and statitistics to bear upon problems of adaptive
and intelligent control.
Some Papers
- Optimal Learning Revisited, AAAI-95 Fall Symposium on Active Learning.es
-
Q-learning for Bandit Problems, Proceedings of Machine Learning 1995.
-
Q-learning for Bandit Problems, CMPSCI Technical Report 95-26,
Department of Computer Science, University of Massachusetts, March 24,
1995 (A slightly longer version--includes simple review of RL).
-
Monte Carlo Matrix Inversion and Reinforcement Learning (with Andrew
Barto), Neural Information Processing Systems -- 6, 1994, pp. 687-694.
-
A Control Variable Perspective for the Optimal Combination
of Truncated Corrected Returns, Unpublished Manuscript, 1994.
-
Reinforcement Learning for Semi-Markov Processes (with Steve
Bradtke), Neural Information Processing Systems -- 7, 1995, pp. ?-?.
-
Solving Bellman's Equation by the Method of Continuation, American
Control Conference, 1994.
- Backpropagation and Bach's 5th Cello Suite (Sarabande), Proceedings
of IEEE Conference on Neural Networks, 1987?.
- Optimal Design of an Electrostatic Lens,
Applied Physics Letters, 1984.
The Adaptive Networks Laboratory.
duff@cs.umass.edu.
Me and my pal Cyrus ("Cyborg")