Simulation of Real Time Problem Solving.
Abstract
Hybrid computer simulations of problem solving can be constructed by combining the production system architecture with a semantic activation process for selecting productions. This technique has been used to construct simulations of problem solving done under time pressure, and in the presence of interruptions. The power of the simulations is limited by the pattern recognition capacity of the semantic networks. This has been analyzed and some mathematical limits obtained. Whether these (or analogous) limits apply to connectionist networks is to base pattern recognition on case-based memory models, which store specific experiences rather than statistical associations between features of experiences. A study is presented showing that case based models may be an alternative to semantic networks. Keywords: Psychology, Connectionism, Problem solving, Information processing.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 30, 1987
- Accession Number
- ADA193436
Entities
People
- Earl Hunt
Organizations
- University of Washington