Intermittent spike train processing through fractional leaky integrate-and-fire neuromorphic unit.

Abstract

The leaky integrate-and-fire (LIF) model provides a fundamental framework for modeling neuronal dynamics in spiking networks. While generalized LIF models can incorporate features, such as spike-frequency adaptation and noise, our study specifically examines its fractional-order extension governed by a relaxation equation with a fractional derivative, whose power-law dynamics emulate long-term memory effects ideal for processing intermittent, scale-invariant signals. Statistical properties of the response of the fractional-order LIF model to a flickering input voltage pulse flow, characterized by a fractional Poisson process of order ν, are evaluated. To implement the fractional LIF model in hardware, we developed a microscale transistor using a network of single-walled carbon nanotubes with an electrolyte gate. The system exhibits fractional-order dynamics, making it well-suited for neuromorphic spiking networks that process scale-invariant signals with long-range temporal correlations.