12/01/1993 09:00 AM
Computer Engineering
This thesis considers issues related to the design of a distributed computing system optimized for research in the area of artificial intelligence applications. We introduce the Associative Processing Environment, a computing model explicitly designed to provide support at the machine level for systems that learn. Issues relating to the implementation of a multicomputer based on this model are discussed with a focus on the development of an effective processor interconnection network. We describe APES, a software tool for the simulation of message traffic in such a system, and present a series of experiments testing the behavior of a parallelized AI application on several topologies. We conclude that the behavior of the application is such that an assumption of uniformly distributed random traffic fails to capture essential aspects of the program\'s communication activity. Notes: M.S. Thesis
