Analog implementations of a binary relation inference network for minimum cost path problems

Minimum cost path problems occur in AI and operations research. A connectionist approach is proposed, based on a binary relation inference network, which demonstrates practical size independence. Continuous-time implementations of the binary relation inference network in solving minimum cost problems are presented. Issues on the design and performance of the various implementations are discussed. It is shown, using commonly available electronic components, that the network is able to provide a correct solution in the time range of microseconds (around 7.5 mu s).<
>