Functional Stability of Software-Hardware Neural Network Implementation The NeuroComp Project

TL;DR

This paper introduces a hardware redundancy approach at the neuron level using microcomputers to enhance neural network stability during operation, differing from traditional dropout regularization.

Contribution

It presents a novel hardware-based method for neural network stability that ensures resilience to failures at the neuron level during deployment.

Findings

Neural networks can maintain functionality despite individual neuron failures.

Hardware redundancy at the neuron level improves system resilience.

The approach differs from dropout by focusing on operational stability rather than training regularization.

Abstract

This paper presents an innovative approach to ensuring functional stability of neural networks through hardware redundancy at the individual neuron level. Unlike the classical Dropout method, which is used during training for regularization purposes, the proposed system ensures resilience to hardware failures during network operation. Each neuron is implemented on a separate microcomputer (ESP32), allowing the system to continue functioning even when individual computational nodes fail.