A biologically plausible network model for pattern storage and recall inspired by Dentate Gyrus

Abstract

In the race to achieve better performance, artificial intelligence has become more about the end rather than the means, which is general intelligence. This work aims to bridge the gap between the two by finding a complementary midline. The objective of this work is to project the role of Dentate Gyrus in enhancing the performance of an autoassociative network, paving the way to develop a biologically plausible neural network which, in the future, would help in simulating the network present in our brain. The proposed network imbibes biological similarities with respect to connectivity, weight updation, and activation function. Dentate Gyrus uses pre-integration lateral inhibition form of learning, and the autoassociative network is implemented using Hopfield network. The performance of the autoassociative network in the presence and absence of Dentate Gyrus is observed across multiple parameters. The results show an increase of 38% in storage capacity and a decrease of 15% in the error tolerance capability of the network in the presence of Dentate Gyrus.