Unsupervised neural network with a memory replacement effect

Abstract

In this paper, we will study the memory replacement effect of a unsupervised learning algorithm for a neural network. The unsupervised learning algorithm on which we base is called learning by experience (LBE). Here, we will modify the network to incorporate a replacement algorithm to increase the learning capability of the network when it faces a memory catastrophe. Simulations to illustrate the feasibility of this effect will be included in this paper.