A pseudo-relaxation learning algorithm for bidirectional associative memory

A fast iterative learning algorithm for the bidirectional associative memory (BAM) called PRLAB is introduced. PRLAB utilizes the pseudo-relaxation method adapted from the relaxation method for solving systems of linear inequalities. PRLAB is very fast, is well suited for a neural network implementation, guarantees the recall of all training patterns, is highly insensitive to learning parameters, and offers high scalability for large applications. PRLAB exploits the maximum storage capacity of the BAM and guarantees perfect recall of all trained pairs. For guaranteed storage, no special form of encoding or preprocessing is necessary