Testing Machine Learning as an Alternative Physics Theory

Nieto-Chaupis H. · International Conference on Engineering and Emerging Technologies, ICEET (2025) · DOI: 10.1109/iceet67911.2025.11424036

Resumen

In this paper, it is demonstrated that the paradigm of machine learning might be a realistic theoretical framework that can reproduced space-time physics equations. Thus, through the usage of artificial neural networks up to three well-known equations are derived. To accomplish this a master equation based in a generalized multilayer perceptron is presented. This has yielded the 1D wave, 1D diffusion and Poisson equations. This proposal clearly might be extended to derive similar equations. The results of paper would suggest that machine learning might be also taken into account as a serious alternative theory by which given a problem in physics, it is then identified a concrete equation as well as its strategy of solution by using numerical approaches in as single perceptron.

Instituciones: Universidad Autónoma del Perú