Implementation Of MNIST Dataset Learning Using Analog Circuit

TL;DR

This paper demonstrates implementing neural networks on the MNIST dataset using analog circuits with capacitors and diodes, controlled by microcontrollers, to analyze real-world performance.

Contribution

It introduces a practical approach to neural network implementation with analog components and microcontrollers, moving beyond simulation-based methods.

Findings

Successful real-world implementation of neural networks with analog circuits

Analysis of circuit-based neural network performance

Use of microcontrollers for controlling analog neural network models

Abstract

There have been many attempts to implement neural networks in the analog circuit. Most of them had a lot of input terms, and most studies implemented neural networks in the analog circuit through a circuit simulation program called Spice to avoid the need to design chips at a high cost and implement circuits directly to input them. In this study, we will implement neural networks using a capacitor and diode and use microcontrollers (Arduino Mega 2560 R3 boards) to drive real-world models and analyze the results.