Some Statistical Analyses of Fitness Landscape Created by the Hopfield Model of Associative Memory
Akira Imada, Keijiro Araki · 1998
We apply evolutionary computations to Hopfield 's neural network model of associative memory. In the model, some of the appropriate configurations of the synaptic weights give a network a function of associative memory. One of our goals is to obtain the distribution of these configurations in the synaptic weight space. For the purpose, we explore a fitness landscape defined on the weight space. Although there have been a fair amount of studies regarding the ruggedness of the fitness landscape, all of them are for discrete genes. In this paper, we apply these methods to our continuous genes which represent the synaptic weights of the Hopfield network. I. Introduction Associative memory is a dynamical system which has a number of stable states with a domain of attraction around them [1]. If the system starts at any state in the domain, it will converge to the stable state. In 1982, Hopfield [2] proposed a fully connected neural network model of associative memory in which information is...