Associative Networks on a Massively Parallel Computer

摘要

A generalization of semantic networks, called an associative network is mapped onto a massively parallel processor which is currently under development. The results show: - The time required to process a query is dependent strictly on the pattern of the query, not on the size of the classes being processed. - The order of processing a query does not affect the speed. - Although we do not receive anywhere near an n-fold speedup by using n processors, we still receive significant performance benefits over a single processor. - The associative network may be used not just as a semantic network, for example, it also allows some problems involving numerical minimizations to be solved efficiently. The primary result of this work is that a large number of simple processors, each responsible for a small piece of information, can work in unison to answer queries significantly faster than a single, highly complex processor can.