Machine Learning at the Flatiron Institute Seminar: Sebastian J. Wetzel

Title: Finding Symmetry Invariants and Conserved Quantities with Artificial Neural Networks

Abstract: In this talk, I will discuss how to find symmetry invariants and conserved quantities with artificial neural networks. This topic falls within the now-emerging subfield of artificial scientific discovery. More precisely, the method is based on interpreting what artificial neural networks learn when trained on data from systems in theoretical physics in order to reveal physical properties. The central method described in this talk is based on Siamese Neural Networks. However, I will also discuss recent developments of methods that are more accurate and allow for learning a full set of independent conserved quantities.