A Novel ANN Structure for Image Recognition

TL;DR

This paper introduces Multi-layer Auto Resonance Networks (ARN), a new neural model for image recognition that is noise-tolerant, tunable, and capable of high accuracy with minimal training data, suitable for edge computing.

Contribution

The paper proposes a novel neural network architecture, ARN, with resonance-based neurons that improve noise tolerance and episodic response, demonstrated on MNIST with high accuracy.

Findings

94% recognition accuracy on MNIST with minimal samples

ARN neurons exhibit noise tolerance and episodic response

Two-layer ARN achieves effective image classification

Abstract

The paper presents Multi-layer Auto Resonance Networks (ARN), a new neural model, for image recognition. Neurons in ARN, called Nodes, latch on to an incoming pattern and resonate when the input is within its 'coverage.' Resonance allows the neuron to be noise tolerant and tunable. Coverage of nodes gives them an ability to approximate the incoming pattern. Its latching characteristics allow it to respond to episodic events without disturbing the existing trained network. These networks are capable of addressing problems in varied fields but have not been sufficiently explored. Implementation of an image classification and identification system using two-layer ARN is discussed in this paper. Recognition accuracy of 94% has been achieved for MNIST dataset with only two layers of neurons and just 50 samples per numeral, making it useful in computing at the edge of cloud infrastructure.