How does a single spiking neuron process information? This question is a long lasting one, which has been constantly posed and pursued by many researchers from different perspectives. In this paper, we tackle this issue from the perspective of reservoir computing using a single Izhikevich neuron as a model system. To prepare reservoir nodes from the response of a single Izhikevich neuron, we used of a technique called time multiplexing, which exploits a time-scale difference between input-output series and the transient dynamics of the single neuron. Based on this scheme, we evaluated the information processing capability of a single Izhikevich neuron using a standard benchmark task. Furthermore, we measured its memory capacity and showed its characteristic memory profile in various parameter settings. Finally, the relationships between the dynamical properties of the Izhikevich neuron and its memory capacity are discussed in detail.

This content is only available as a PDF.