Learning Conservation Principles in Particle Physics
by Oliver Schulte, Simon Fraser University
This talk presents machine learning algorithms for data analysis in particle
physics. One of the major goals of particle physics is to find conservation
principles that predict which interactions among elementary particles are
possible; well-known examples include the conservation of energy and electric
charge. I describe a program that performs a systematic search for conserved
quantities over a set of observed particle reactions.
On current data, the program finds a set of conservation principles that are equivalent to those in standard particle theory, in the sense that the program classifies reactions as "possible" or "impossible" exactly as the standard laws do. If we add the specification that conservation principles should correspond to particle clusters, the program rediscovers exactly the standard laws (this is joint work with Mark Drew). The algorithm also has the capability to introduce new particles when this produces a better fit with the data. This leads to the computation of a novel experiment for testing the hypothesis that the neutrino is a Majorana particle, which is one of the crucial questions in current particle physics. The talk is self-contained and assumes no previous knowledge of particle physics.