Logic control strategies for parallel/distributed intelligent systems
Abstract
Knowledge-based problem solving and more specifically expert problem solving has become one of the most successful areas of Artificial Intelligence in recent years. Therefore, its methodology has been applied to a variety of fields including very computationally expensive ones such as machine vision. In order to increase the execution speed of such large applications, it becomes natural then to implement knowledge-based systems on multiprocessors. In this thesis a study of the possibilities of parallelism in expert systems is undertaken. First, a survey of parallelism in problem solving is done with the aim of applying those techniques to the more restricted paradigm of expert systems. The parallelism is analyzed at the levels of architectural support languages and tools, and heuristics. This thesis examines the performance of several control strategies for Expert Systems in multiprocessor machines, and proposes SIMPL, a language model for simultaneous execution of knowledge-based programs. The model is based in logic programming and permits the user to parallelize existing Prolog software by means of a number of control strategies included as meta-rules. Contributions of this work lie in the areas of symbolic parallel programming techniques and knowledge engineering.
Degree
Ph.D.
Advisors
Jamieson, Purdue University.
Subject Area
Electrical engineering|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.