Machine learning in particle physics - going beyond classification and regression
by
Michael Spannowsky
→
Europe/Berlin
313 (MPI Meeting rooms)
313
MPI Meeting rooms
Description
Machine learning methods have become a standard tool to classify events according to their probability to be induced by competing hypotheses. Another standard application of such techniques is regression, e.g. for the purpose of unfolding detector effects in the reconstruction of measured objects. Neural networks, however, are not limited to exploring correlations in pre-existing data, but also provide a powerful and rather generic framework for the numerical solution of functionals, e.g. differential or integral equations. With an eye on existing challenges, I will briefly discuss various ways in which neural networks are commonly used, and will then explain how they can be applied to solving differential equations. The latter will be applied to the study of phase transitions and gravitational wave spectra.
