Chess-playing programs and the problem of complexity
Computers & thought
Realization of a geometry-theorem proving machine
Computers & thought
Algebraic simplification: a guide for the perplexed
Communications of the ACM
Syntax-directed interpretation of classes of pictures
Communications of the ACM
ACM '72 Proceedings of the ACM annual conference - Volume 1
Perceptrons: An Introduction to Computational Geometry
Perceptrons: An Introduction to Computational Geometry
Problem-Solving Methods in Artificial Intelligence
Problem-Solving Methods in Artificial Intelligence
| Hi-index | 0.00 |
This paper discusses the “standard” techniques used by problem solving and pattern recognition programs. It is pointed out that evaluation functions (often called discriminant functions) lie at the heart of these programs. Simplicity appears to be an important property of evaluation functions because evaluation functions which are both relatively accurate and efficient are in some sense simple. In addition, simple evaluation functions are easier to learn because there are fewer parameters to estimate. The real difficulty is that no general method exists which extracts a good set of features for a particular problem. Even if this were possible one is still faced with the task of combining these features to obtain the “answer”. In the case of pattern recognition this combination takes the form of connectives such as arithmetic operations or Boolean operations. However, in the case of problem solving, things are much more complicated since features are at the bottom of a search procedure which looks at many different problem states. This presents the difficulty of making good use of such a large volume of information.