Martin Mueller, University of Alberta
Associate Professor
Computing Science
Edmonton, Alberta
mmueller@cs.ualberta.ca
Office:
(780) 492-3703
Expert In
Bio/Research
I am interested in two aspects of powerful search-based programs: search algorithms for both single-agent search and two player games, and knowledge engineering for building evaluation functions and search control.
A common theme of my recent research is "search algorithms that use more ...
Click to Expand >>
A common theme of my recent research is "search algorithms that use more ...
Click to Expand >>
Bio/Research
I am interested in two aspects of powerful search-based programs: search algorithms for both single-agent search and two player games, and knowledge engineering for building evaluation functions and search control.
A common theme of my recent research is "search algorithms that use more than just numbers". In particular, I have developed three methods called decomposition search, partial order bounding and proof-set search. Decomposition search is a problem decomposition approach to solving games that uses combinatorial game techniques. Partial order bounding is a method that can model incomplete knowledge in game evaluation functions in a principled way, and proof-set search is a refinement of Victor Allis' proof-number search.
The development of these search algorithms was motivated by practical experience building game-playing programs, especially for the game of Go. This very difficult game poses many challenges which expose the limitations of classical game-playing methods. The aim of my research is to develop new theoretical and practical approaches to overcome these problems.
Click to Shrink <<
A common theme of my recent research is "search algorithms that use more than just numbers". In particular, I have developed three methods called decomposition search, partial order bounding and proof-set search. Decomposition search is a problem decomposition approach to solving games that uses combinatorial game techniques. Partial order bounding is a method that can model incomplete knowledge in game evaluation functions in a principled way, and proof-set search is a refinement of Victor Allis' proof-number search.
The development of these search algorithms was motivated by practical experience building game-playing programs, especially for the game of Go. This very difficult game poses many challenges which expose the limitations of classical game-playing methods. The aim of my research is to develop new theoretical and practical approaches to overcome these problems.
Click to Shrink <<