Balanced Resonate-and-Fire Neurons

TL;DR

The paper introduces the balanced resonate-and-fire (BRF) neuron model, which improves learning efficiency, reduces spiking activity, and enhances training stability in recurrent spiking neural networks for sequence tasks.

Contribution

The authors propose the balanced RF neuron model that overcomes limitations of traditional RF neurons, demonstrating improved performance and efficiency in RSNNs.

Findings

BRF neurons achieve higher task performance

Networks with BRF neurons produce fewer spikes

BRF-RSNNs train faster and more stably

Abstract

The resonate-and-fire (RF) neuron, introduced over two decades ago, is a simple, efficient, yet biologically plausible spiking neuron model, which can extract frequency patterns within the time domain due to its resonating membrane dynamics. However, previous RF formulations suffer from intrinsic shortcomings that limit effective learning and prevent exploiting the principled advantage of RF neurons. Here, we introduce the balanced RF (BRF) neuron, which alleviates some of the intrinsic limitations of vanilla RF neurons and demonstrates its effectiveness within recurrent spiking neural networks (RSNNs) on various sequence learning tasks. We show that networks of BRF neurons achieve overall higher task performance, produce only a fraction of the spikes, and require significantly fewer parameters as compared to modern RSNNs. Moreover, BRF-RSNN consistently provide much faster and more stable training convergence, even when bridging many hundreds of time steps during backpropagation through time (BPTT). These results underscore that our BRF-RSNN is a strong candidate for future large-scale RSNN architectures, further lines of research in SNN methodology, and more efficient hardware implementations.