Recognition of handwritten MNIST digits on low-memory 2 Kb RAM Arduino board using LogNNet reservoir neural network

TL;DR

This paper presents a compact LogNNet reservoir neural network algorithm capable of recognizing MNIST handwritten digits on a low-memory 2 Kb Arduino microcontroller with 82% accuracy, suitable for IoT and edge applications.

Contribution

It introduces a memory-efficient neural network algorithm that operates on minimal hardware without data pre-processing, enabling AI on constrained IoT devices.

Findings

Achieved 82% recognition accuracy on MNIST digits.

Demonstrated operation on 2 Kb RAM Arduino microcontroller.

Showed the algorithm's suitability for edge computing and medical biosensors.

Abstract

The presented compact algorithm for recognizing handwritten digits of the MNIST database, created on the LogNNet reservoir neural network, reaches the recognition accuracy of 82%. The algorithm was tested on a low-memory Arduino board with 2 Kb static RAM low-power microcontroller. The dependences of the accuracy and time of image recognition on the number of neurons in the reservoir have been investigated. The memory allocation demonstrates that the algorithm stores all the necessary information in RAM without using additional data storage, and operates with original images without preliminary processing. The simple structure of the algorithm, with appropriate training, can be adapted for wide practical application, for example, for creating mobile biosensors for early diagnosis of adverse events in medicine. The study results are important for the implementation of artificial intelligence on peripheral constrained IoT devices and for edge computing.