Spiking Continuous Attractor Neural Networks with Spike Frequency Adaptation for Anticipative Tracking

Abstract

Continuous attractor neural network (CANN) is a canonical model for neural information representation and processing, which has been applied to describe the encoding of continuous features, such as orientation, head direction and spatial location in neural systems. Specifically, theoretical studies based on a firing-rate model have found that a CANN with negative feedback, such as spike frequency adaptation (SFA), has the capability of tracking a continuously moving stimulus anticipatively. In this study, facing the booming development of neuromorphic computing using spiking neural networks (SNNs), we built a spiking continuous attractor neural network (S-CANN) with SFA to implement anticipative tracking. Further, we simplified the model, in terms of connection weights, external inputs, and network size, to facilitate its implementation with neuromorphic hardware.