Eric Wefald

Home * People * Eric Wefald

Eric Huang Wefald, (died August 31, 1989)
was an American computer scientist and researcher from University of California, Berkeley, California, who co-authored with Stuart Russell on multiple papers on search control, machine learning, and metareasoning, that is the process of reasoning about reasoning itself. Eric Wefald obtained a Master of philosophy from Princeton in 1985, and was completing a doctorate in artificial intelligence at U.C. Berkeley, when he died in a car accident near Bordeaux, France, on August 31, 1989.

Decision-Theoretic Search Control

Abstract from Stuart Russell and Eric Wefald (1988). Decision-Theoretic Search Control: General Theory and an Application to Game-Playing. :

In this paper we outline a general approach to the study of problem-solving, in which search steps are considered decisions in the same sense as actions in the world. Unlike other metrics in the literature, the value of a search step is defined as a real utility rather than as a quasi-utility, and can therefore be computed directly from a model of the base-level problem-solver. We develop a formula for the value of a search step in a game-playing context using the single-step assumption, namely that a computation step can be evaluated as it was the last to be taken. We prove some meta-level theorems that enable the development of a low- overhead algorithm, MGSS, that chooses search steps in order of highest estimated utility. Although we show that the single-step assumption is untenable in general, a program implemented for the game of Othello appears to rival an alpha-beta search with equal node allocations or time allocations. Pruning and search termination subsumes or improve on many other algorithms. Single-agent search, as in the A algorithm, yields a simpler analysis, and we are currently investigating applications of the algorithm developed for this case. 

Optimal Game-Tree Search

Abstract from Stuart Russell and Eric Wefald (1989). On optimal game-tree search using rational metareasoning :

In this paper we outline a general approach to the study of problem-solving, in which search steps are considered decisions in the same sense as actions in the world. Unlike other metrics in the literature, the value of a search step is defined as a real utility rather than as a quasi-utility, and can therefore be computed directly from a model of the base-level problem-solver. We develop a formula for the expected value of a search step in a game-playing context using the single-step assumption, namely that a computation step can be evaluated as it was the last to be taken. We prove some meta-level theorems that enable the development of a low-overhead algorithm, MGSS*, that chooses search steps in order of highest estimated utility. Although we show that the single-step assumption is untenable in general, a program implemented for the game of Othello soundly beats an alpha-beta search while expanding significantly fewer nodes, even though both programs use the same evaluation function. 

Selected Publications

• Stuart Russell, Eric Wefald (1988). Decision-Theoretic Search Control: General Theory and an Application to Game-Playing. CS Technical Report 88/435, University of California, Berkeley
• Stuart Russell, Eric Wefald (1988). Multi-Level Decision-Theoretic Search. AAAI Symposium on Computer Game-Playing, Stanford.
• Stuart Russell, Eric Wefald (1989). On optimal game-tree search using rational metareasoning. IJCAI 1989
• Eric Wefald, Stuart Russell (1989). Adaptive Learning of Decision-Theoretic Search Control Knowledge. 6th International Workshop on Machine Learning
• Stuart Russell, Eric Wefald (1991). Principles of Metareasoning. Artificial Intelligence, Vol. 49, Nos. 1-3
• Stuart Russell, Eric Wefald (1991). Do the right thing: studies in limited rationality. MIT Press

References

Up one level