Six networks on a universal neuromorphic computing substrate

TL;DR

This paper introduces a highly configurable mixed-signal neuromorphic computing substrate capable of emulating diverse neural networks with high speed and flexibility, facilitating neuroscientific research and network experimentation.

Contribution

It presents a universal neuromorphic hardware platform that supports various network topologies and parameters, with calibration and user-friendly tools for neuroscientists.

Findings

Successfully emulated six different neural networks

Achieved high acceleration compared to conventional computers

Reduced analog noise through calibration routines

Abstract

In this study, we present a highly configurable neuromorphic computing substrate and use it for emulating several types of neural networks. At the heart of this system lies a mixed-signal chip, with analog implementations of neurons and synapses and digital transmission of action potentials. Major advantages of this emulation device, which has been explicitly designed as a universal neural network emulator, are its inherent parallelism and high acceleration factor compared to conventional computers. Its configurability allows the realization of almost arbitrary network topologies and the use of widely varied neuronal and synaptic parameters. Fixed-pattern noise inherent to analog circuitry is reduced by calibration routines. An integrated development environment allows neuroscientists to operate the device without any prior knowledge of neuromorphic circuit design. As a showcase for the capabilities of the system, we describe the successful emulation of six different neural networks which cover a broad spectrum of both structure and functionality.