Working Memory in a Recurrent Spiking Neural Networks With Heterogeneous Synaptic Delays

TL;DR

This paper introduces a recurrent spiking neural network with heterogeneous synaptic delays trained via surrogate-gradient backpropagation, effectively storing and recalling multiple temporal spike patterns for energy-efficient neuromorphic applications.

Contribution

The authors propose a novel SNN architecture with multiple synaptic delays trained end-to-end, demonstrating high accuracy in storing complex temporal patterns.

Findings

Achieved a mean F1 score of 1.0 on a synthetic benchmark with 16 patterns.

Heterogeneous delays enable efficient working memory in SNNs.

Recall propagates forward from the initial window, showing temporal pattern storage.

Abstract

Working memory -- the ability to store and recall precise temporal patterns of neural activity -- remains an open challenge for spiking neural networks (SNNs). We propose a recurrent SNN of $N$ neurons in which each synapse is equipped with $D = 41$ delays, modelled as a weight tensor $\mathbf{W} \in \mathbb{R}^{N \times N \times D}$ and trained end-to-end with surrogate-gradient backpropagation through time. The network stores $M$ arbitrary target spike patterns by representing each as a sequential chain of overlapping Spiking Motifs: contiguous windows of length $D$ that uniquely predict spikes at the next time step. On a synthetic benchmark of $M=16$ patterns ($N=512$ neurons, $T=1000$ steps), training achieves a mean F1 score of $1.0$, with recall emerging first near the clamped initialisation window and propagating forward in time. This result demonstrates that heterogeneous delays provide an efficient substrate for working memory in SNNs, enabling energy-efficient neuromorphic edge deployment.