hxtorch.snn: Machine-learning-inspired Spiking Neural Network Modeling on BrainScaleS-2

TL;DR

This paper introduces hxtorch.snn, a machine learning framework that simplifies the design, training, and simulation of spiking neural networks on BrainScaleS-2 hardware, supporting full automation and hardware-in-the-loop workflows.

Contribution

It presents a novel, fully-automated, machine learning-based modeling framework for BrainScaleS-2 that supports hardware-in-the-loop training and seamless hardware-software transitions.

Findings

Successful classification on Yin-Yang dataset

Supports auto differentiation and hardware-in-the-loop training

Enables transition between hardware emulation and software simulation

Abstract

Neuromorphic systems require user-friendly software to support the design and optimization of experiments. In this work, we address this need by presenting our development of a machine learning-based modeling framework for the BrainScaleS-2 neuromorphic system. This work represents an improvement over previous efforts, which either focused on the matrix-multiplication mode of BrainScaleS-2 or lacked full automation. Our framework, called hxtorch.snn, enables the hardware-in-the-loop training of spiking neural networks within PyTorch, including support for auto differentiation in a fully-automated hardware experiment workflow. In addition, hxtorch.snn facilitates seamless transitions between emulating on hardware and simulating in software. We demonstrate the capabilities of hxtorch.snn on a classification task using the Yin-Yang dataset employing a gradient-based approach with surrogate gradients and densely sampled membrane observations from the BrainScaleS-2 hardware system.