Better physics-informed machine learning for better learning about physics

Abstract:

Motivated by the large volume and high complexity of experimental data and mathematical structures, particle physics has a long tradition of employing state of the art computing and analysis techniques. Recent progress in machine learning and artificial intelligence have further pushed this trend, and these approaches are now ubiquitous in our field.
This colloquium will focus on two questions: How to improve machine learning architectures by combining raw computing power with insights from physics?
And which new ways of gaining insight from physics data could be unlocked by modern machine learning?