Study of associative processing techniques

Abstract : This report describes designs of two novel stored program digital machines which process non-numeric data by use of distributed associative logic. The first machine is an associative processor, which is programmed to abstract properties from visual or other patterns, and to classify patterns by use of these properties. Associative logic simultaneously evaluates many linear threshold decision functions at high solution rates and with high digital accuracy. Functional behavior of the machine is easily modified by changing memory contents. The techniques used to evaluate threshold functions may also be used to evaluate arbitrary Boolean functions of stored and external variables. The second machine described is an associative file processor. This processor rapidly searches a data base, simultaneously against many descriptors. The processor uses an inexpensive disc memory for record storage. It allows variable record lengths with simple insertion, deletion, or modification of stored records. It allows equality research, limit search, or between limits search on all descriptors within input queries. Block diagrams, command sets, and logic descriptions are given for each processor. Each processor is programmed to accomplish its intended tasks. Applications for each processor are described. (Author)