Some Short Examples Mapping Algorithms onto Networks

This chapter reviews a variety of problems for which networks would appear to be attractive. However, they do not have an obvious mapping onto some simple structure, in the way that near-neighbor perceptual processes map onto single instruction, multiple data (SIMD) arrays. Some are the problems that people have scarcely begun to formulate for networks. Therefore, these should be read for what they are—rough sketches that almost certainly would change beyond recognition during the process of precise formulation and programming for a network. Both a breadth-first and a heuristic-search program might be executed on an SIMD network and by a multiple instruction, multiple data network. The chapter also discusses searches for question-answering and database access. It presents searches through semantic memory nets. The straightforward way to handle semantic memory searches would be to assign each node in the semantic memory network to a different processor, with a link between processors wherever there is a link between semantic memory nodes.