Accelerating Machine Learning via the Weber-Fechner Law

TL;DR

This paper proposes applying the Weber-Fechner Law to neural networks, demonstrating that incorporating perceptual scaling can significantly accelerate learning and improve accuracy on the MNIST dataset with limited resources.

Contribution

It introduces a novel method of integrating the Weber-Fechner Law into neural networks to enhance learning speed and efficiency for human concept recognition.

Findings

Rapid training convergence on MNIST dataset

Improved accuracy with limited computational resources

Surprising performance gains from perceptual scaling

Abstract

The Weber-Fechner Law observes that human perception scales as the logarithm of the stimulus. We argue that learning algorithms for human concepts could benefit from the Weber-Fechner Law. Specifically, we impose Weber-Fechner on simple neural networks, with or without convolution, via the logarithmic power series of their sorted output. Our experiments show surprising performance and accuracy on the MNIST data set within a few training iterations and limited computational resources, suggesting that Weber-Fechner can accelerate machine learning of human concepts.