Abstract

Associative memory (AM) is appropriate for storage of certain types of information. Conventional location addressable memory (LAM) can more efficiently store deterministic spatial and temporal patterns. A problem solving system may combine these types of memory under control of a microprogram, or context, sequencer. The solution of a problem progresses through a sequence of steps. Some steps involve decisions that refine or redirect the context of the solution process. The LAM is used to store meaningful sequences of solution steps. Output of the AM contains information associated with the input data, including the direction of the next appropriate step in the solution process. This information is used by the context sequencer to determine whether the next sequential step of the current solution path will be used or if a branch should be taken. Output of the sequencer indexes into the LAM for the next step information. Applications of sequential associative processing are presented. Implementation and integration of the component parts, the AM, the LAM, and the context sequencer are discussed.

© 1985 Optical Society of America