Avoiding Catastrophic Forgetting by a Biologically Inspired Dual-Network Memory Model

In neural networks, when new patterns are learned by a network, they radically interfere with previously stored patterns. This drawback is called catastrophic forgetting. In this paper, we propose a biologically inspired dual-network memory model which can much reduce this forgetting. The proposed model consists of two distinct neural networks: a hippocampal network and a neocortical network. Information given is first stored in a hippocampal network and then it is transferred to the neocortical network. The chaotic recall of the hippocampal network enables the extraction of original patterns, and owing to this, they can be interleaved with previous information stored in the neocortical network. Thus, catastrophic forgetting can be avoided. Computer simulation results show the effectiveness of the proposed dual-network memory model.