Presented by: Keith Ulmer, University of Colorado Boulder

Abstract: The Large Hadron Collider at CERN probes the fundamental structure of nature at the highest energies ever achieved in a laboratory. By searching the enormous number of particle collisions produced at the LHC, scientists seek answers to fundamental questions including the nature of dark matter, the origin of mass, and whether the forces of nature can be unified. But how do we search for new physics when we don't know what the new physics looks like?

This talk explores a new approach being developed by the CMS experiment: using machine-learning-based anomaly detection directly in the detector's real-time data acquisition system to identify unusual collisions without targeting a specific signal. Rather than deciding in advance which signatures are interesting, this approach allows CMS to search for unexpected patterns in the data itself. I will discuss the physics motivation and development of the AXOL1TL anomaly detection algorithm, the technical challenges of deploying advanced machine learning in hardware systems processing roughly 50 terabytes of data per second, and how this new capability could open a novel window for discovering physics beyond the Standard Model at the LHC.

Host: Tobin Munsat