A comparative analysis of a neural network with calculated weights and a neural network with random generation of weights based on the training dataset size

TL;DR

This paper compares neural networks with analytically calculated weights to those with randomly initialized weights, demonstrating that pre-calculated weights lead to faster training and greater robustness on the MNIST dataset.

Contribution

It introduces a method of calculating neural network weights analytically and compares its performance to random initialization across different dataset sizes.

Findings

Pre-calculated weights enable faster training.

Pre-calculated weights improve robustness to smaller datasets.

Analytical weight calculation enhances training efficiency.

Abstract

The paper discusses the capabilities of multilayer perceptron neural networks implementing metric recognition methods, for which the values of the weights are calculated analytically by formulas. Comparative experiments in training a neural network with pre-calculated weights and with random initialization of weights on different sizes of the MNIST training dataset are carried out. The results of the experiments show that a multilayer perceptron with pre-calculated weights can be trained much faster and is much more robust to the reduction of the training dataset.