Including STDP to eligibility propagation in multi-layer recurrent spiking neural networks

TL;DR

This paper investigates the integration of STDP into eligibility propagation for training multi-layer recurrent spiking neural networks, demonstrating performance improvements and insights into network architecture effects.

Contribution

It introduces the inclusion of STDP-like behavior into e-prop for SNNs and compares single-layer versus multi-layer architectures, revealing performance benefits.

Findings

Including STDP improves classification with ALIF neurons.

STDP does not enhance Izhikevich neuron performance.

Single-layer e-prop outperforms multi-layer configurations.

Abstract

Spiking neural networks (SNNs) in neuromorphic systems are more energy efficient compared to deep learning-based methods, but there is no clear competitive learning algorithm for training such SNNs. Eligibility propagation (e-prop) offers an efficient and biologically plausible way to train competitive recurrent SNNs in low-power neuromorphic hardware. In this report, previous performance of e-prop on a speech classification task is reproduced, and the effects of including STDP-like behavior are analyzed. Including STDP to the ALIF neuron model improves the classification performance, but this is not the case for the Izhikevich e-prop neuron. Finally, it was found that e-prop implemented in a single-layer recurrent SNN consistently outperforms a multi-layer variant.