|
|
|
 
Associate Professor and Chair of Computer Science at Gettysburg
College |
|
|
| COURSES |
|
Previous courses:
CS 103: Introduction
to Computing (Home Page, Online
Text),
CS
112: Computer Science II,
CS 216: Data
Structures,
CS 341: Programming
Languages,
CS 371: Artificial
Intelligence,
CS 374: Compilers,
CS
391: Selected Topics:
Data Mining,
CS
391: Selected Topics:
Machine Learning,
CS 392:
Selected Topics: Game
Artificial Intelligence.
|
| RESOURCES |
|
|
| GAMES AND PUZZLES |
|
|
|
ABOUT |
|
Todd W. Neller is an Associate Professor and
Chair of
Computer Science at Gettysburg College. A Cornell University Merrill
Presidential Scholar, he received a B.S. in Computer Science with
distinction in 1993. In 2000, he received his Ph.D. with distinction
in teaching at Stanford University, where he was awarded a Stanford
University Lieberman Fellowship, and the George E. Forsythe Memorial
Award for excellence in teaching. His dissertation concerned
extensions of artificial intelligence (AI) search algorithms to hybrid
dynamical systems, and the refutation of hybrid system properties
through simulation and information-based optimization. A game
enthusiast, Neller has in recent years enjoyed pursuing game AI
challenges, computing optimal play for jeopardy dice games such as
Pass the Pigs, new reasoning algorithms for Clue/Cluedo, and optimal
Risk attack and defense policies.
|
|
| CLUE® /
CLUEDO® |
MYSTERY OF THE GAME: What would it mean to play the board
game Clue perfectly? Far more than a children's game,
optimal Clue play combines elements of constraint satisfaction,
reasoning about knowledge, bluff, and other facets of game
theory.
Our
current research investigates approximations of optimal Clue
play. In particular, we are examining tradeoffs of speed and
accuracy for probabilistic estimation.
|
| REINFORCEMENT LEARNING |
| REINFORCEMENT LEARNING (RL) techniques allow computers to
learn good behaviors through trial and error rather than have such behaviors explicitly programmed. We apply such techniques to the
metalevel control of search and optimization algorithms. For
example, we have successfully applied RL techniques to the control
of simulated annealing, dynamically adjusting the temperature and
deciding when to terminate optimization.
| |
