A BIDIRECTIONAL VECTOR ASSOCIATIVE MEMORY ARCHITECTURE WITH APPLICATION TO NEURAL NETWORKS

A bidirectional architecture for associative memory (AM) capable of ve ctor arithmetic operations is proposed. By introducing a pair of maski ng and tagging mechanisms, the conventional concepts of bit-operations and word-operations in AM have been generalized to row and column ope rations, respectively. The proposed architecture demonstrates a symmet rical functionality such that associative processing can be performed in both column and row directions. A set of built-in vector arithmetic and logic units (VALU) is designed to perform the basic vector operat ions, which offers O(max{n, m}) speed-up for vector operations over th e conventional AM at an O(n + m) cost for realizing an n x m AM array. As an applicational example, an associative processing implementation for artificial neural networks is presented. Copyright (C) 1996 Elsev ier Science Ltd.