IIHE invited seminar: To be rescheduled: Adversarial Games for Particle Physics

by Dr Gilles Louppe (Université de Liège)

Wednesday 28 Feb 2018, 14:00 → 15:00 Europe/Brussels

1G012 (Universe)

In this talk, we will discuss how recent advances in adversarial neural network training can be used to solve fundamental problems in particle physics. We will first introduce generative adversarial networks (GANs) and present some state-of-the-art results from the field of machine learning. We will then show how GANs can be used for learning fast but accurate approximations of scientific simulators. Then, we will show how adversarial training can be used for building predictive models that are insensitive to systematic uncertainties. Finally, the last part of the talk will address the problem of likelihood-free inference through adversarial variational optimization.