Stochastic Equilibrium Propagation for Spiking Convergent Recurrent Neural Networks

TMLR Paper8829 Authors

Abstract: Spiking Neural Networks (SNNs) promise energy-efficient, sparse, biologically inspired computation. Training them with Backpropagation Through Time (BPTT) and surrogate gradients achieves strong performance but remains biologically implausible. Equilibrium Propagation (EP) provides a more local and biologically grounded alternative. However, existing EP frameworks for SNNs largely rely on deterministic neurons, which require complex mechanisms to handle spiking discontinuities and do not scale beyond simple benchmarks such as MNIST and Fashion-MNIST. Inspired by the stochastic nature of biological spiking mechanism and recent hardware trends, we propose a stochastic EP framework that integrates probabilistic spiking neurons into the EP paradigm. This formulation smoothens the optimization landscape, stabilizes training, and enables efficient and scalable learning in SNNs. We provide theoretical guarantees showing that the proposed stochastic EP dynamics approximate deterministic EP under mean-field theory, thereby inheriting its underlying theoretical guarantees. The proposed framework narrows the performance gap with both BPTT-trained SNNs and EP-trained non-spiking convergent recurrent neural networks (CRNNs) on CIFAR-10, DVS Gesture, and IMDB datasets, while preserving temporal and spatial locality. Our results highlight stochastic EP as a promising approach for neuromorphic and on-chip learning.

Submission Type: Regular submission (no more than 12 pages of main content)

Assigned Action Editor: ~Shi_Gu1

Submission Number: 8829