Representation of Time-Series by a Self-Similar Set in a Model of Hippocampal CA1

Abstract

Because episodic memory includes a time series of events, an underlying dynamics for the formation of episodic memory is considered to employ a mechanism of encoding sequences of events. The “Cantor coding” hypothesis in hippocampal CA1 has been proposed, which provides a scheme for encoding temporal sequences of events. Here, in order for investigating the Cantor coding in detail, we constructed a model for the CA1 network which consists of conductance-based model neurons. It was assumed that the CA3 outputs temporal sequences of spatial patterns to CA1. It was shown that the output patterns of CA1 were hierarchically clustered in a self-similar manner according to the similarity of input time series. The dependency of the efficacy of encoding on the input time interval and its robustness against noise was investigated.