Mixed-signal implementation of feedback-control optimizer for single-layer Spiking Neural Networks

TL;DR

This paper demonstrates a mixed-signal neuromorphic processor implementing feedback-control optimizers for on-chip training of single-layer spiking neural networks, achieving performance comparable to simulations and gradient methods.

Contribution

It presents the first proof-of-concept hardware implementation of feedback-control optimizers for neuromorphic systems, enabling on-chip learning under realistic mixed-signal constraints.

Findings

On-chip training matches simulation and baseline performance.

Feedback-control optimizers are feasible in mixed-signal neuromorphic hardware.

The approach supports autonomous, adaptive learning in neuromorphic devices.

Abstract

On-chip learning is key to scalable and adaptive neuromorphic systems, yet existing training methods are either difficult to implement in hardware or overly restrictive. However, recent studies show that feedback-control optimizers can enable expressive, on-chip training of neuromorphic devices. In this work, we present a proof-of-concept implementation of such feedback-control optimizers on a mixed-signal neuromorphic processor. We assess the proposed approach in an In-The-Loop(ITL) training setup on both a binary classification task and the nonlinear Yin-Yang problem, demonstrating on-chip training that matches the performance of numerical simulations and gradient-based baselines. Our results highlight the feasibility of feedback-driven, online learning under realistic mixed-signal constraints, and represent a co-design approach toward embedding such rules directly in silicon for autonomous and adaptive neuromorphic computing.