Machine Learning Plays the Role of Classical Mechanics

Nieto-Chaupis H. · 2026 IEEE/ACIS 24th International Conference on Software Engineering Research, Management and Applications, SERA 2026 (2026) · DOI: 10.1109/sera69989.2026.11618624

Resumen

If physics equations are based in differential equations as well as integrations, then one can argue that any parallel theory based on these mathematical tools can replace classical physics by using the formalism as a fixed backbone. For instance, consider the case of classical mechanics, all of them might be replaced by pseudo equations that are generated by artificial neural networks. In this paper, it is presented the fact that while by using perceptron, one can retrieve the main equations of classical mechanics in a straightforward manner without inconsistencies. To accomplish this, a model of perceptron and a null activation function is proposed as potential generator of equations belonging to classical physics, specifically. The results of paper, would suggest the fact that machine learning might to recover well-known physics equations, as well as to be a tool to discover unseen physics principles or laws.

Instituciones: Universidad Autónoma del Perú