Approach to Optimization Problems Using Artificial Neural Networks

Abstract

In this article, we aim to provide an understanding of how optimization problems can be approximated using the dynamics of artificial neural networks, specifically Hopfield networks. We have chosen Hopfield networks for their suitability in achieving our goal: the online resolution of linear, quadratic, nonlinear, and combinatorial optimization problems. We construct, develop, and formally justify an integral methodology that allows us to address various types of optimization problems and build a Hopfield neural network that computationally and formally solves them, comparing the results with more recent publications