UNLEARNING ALGORITHM IN ASSOCIATIVE MEMORY

In the present brief we incorporate into a synthesis procedure for a c lass of discrete-time neural networks unlearning capability. The propo sed technique increases storing capacity while maximizing the domain o f attraction of each desired pattern to be stored. Making use of learn ing and forgetting capabilities, neural networks generated by the meth od advanced herein are capable of learning new patterns as well as for getting learned patterns without the necessity of recomputing the enti re interconnection weights and external inputs. The unlearning algorit hm developed is then utilized off-lineto equalize the basins of attrac tion for each desired pattern to be stored, and to minimize the number of spurious states. Specific examples are given to illustrate the str engths and weaknesses of the methodology advocated herein.