A minimum interconnection direct storage model of a neural bidirectional memory

This paper proposes an efficient and improved model of a direct storage bidirectional memory, IBAM, which directly stores the X and Y, associated sets of M bipolar binary vectors, and does not store the complementary memories. It requires O(N) or about 15% of interconnections with weight strength ranging between plusmn1, and is computationally very efficient as compared to sequential, intra-connected, and other models of BAMs of outer-product type. The effect of concatenation of vectors being stored, retrieval constraints and orthogonality issues and restrictions on the length, in bits, and number of vectors to be stored are discussed. It is simple and robust in structure, VLSI realizable, modular and expandable neural network bidirectional associative memory model as the addition or deletion of a pair of vectors does not require changes in the strength of interconnections of the entire memory matrix. The stability analysis of the proposed model has been carried out, which shows its superior performance, faster convergence and retrieval time, when compared to traditional sequential and intra-connected bidirectional memories. The analysis of signal to noise ratio, storage capacity, and performance of the proposed model has been carried out. Its performance has been demonstrated by means of numericals.