Evolving weight matrices to increase the capacity of Hopfield neural network associative memory using hybrid evolutionary algorithm

This paper describes the implementation of a hybrid evolutionary technique to increase the capacity of associative memory in Hopfield type of neural network. Various operators of genetic algorithm (mutation, crossover, elitism etc) are used to evolve the population of optimal weight matrices for the purpose of recall of the prototype input patterns with induced noise. The optimal weight matrix found during the training is used as seed for starting the GA, instead starting with random weight matrix. It has been observed that for Hopfield neural networks of various sizes the recalling is successful if number of patterns stored is within 40% of the total number of nodes in the network which is towards the higher side than the earlier reported capacity.