TRANSPUTER-BASED IMPLEMENTATION OF DISTRIBUTED ASSOCIATIVE MEMORIES

The Distributed Associative Memory (DAM) model operates as a basic dev ice inside an image-classification system, whose accuracy is increased by the restoration ability of the associative mechanism. The paper de scribes a methodology for implementing the DAM model on transputer-bas ed hierarchical parallel architectures. First, a theoretical analysis reformulates the basic memory model to fit a practical implementation; then, the parallel system's efficiency is defined analytically. Exper imental results on a significant testbed confirm the classifier's effe ctiveness and the consistency of theoretical predictions. (C) 1997 Els evier Science B.V.