Machine Learning for Fundamental Physics Discovery with High Resolution Particle Imaging Detectors

In search for increasingly rarer signatures of “new physics”, today’s particle detectors are streaming raw data with increasingly higher resolution, volume, and complexity, posing growing challenges to data processing and physics data analysis. At the same time, recent advances in machine learning and artificial intelligence provide an unprecedented opportunity to dig deeper and extract more information out of large data sets. The talk will present the case of a particular detector technology commonly used in neutrino physics, characterized by exorbitant amounts of image-like raw data and unprecedented demands for real-time data processing, where "accelerated machine learning" can be used to extract rare signatures, enabling fundamental physics discoveries.

Biography:

Georgia Karagiorgi is an Associate Professor of Physics at Columbia University. She received her Ph.D. in Experimental High Energy Physics from MIT in 2010. Her main research is in experimental particle physics, including searches for new physics in the neutrino sector with particle-accelerator-based and/or large underground neutrino experiments. The discovery and study of rare processes with the next generation of neutrino detectors requires continuous and efficient processing of data streams with rates of up to multiple terabytes per second. Her current research includes data processing hardware design and exploration of digital data processing techniques for high data rate environments, and more recently she has been exploring the implementation of machine learning algorithms on field-programmable gate arrays. She is currently serving as Technical Lead for the data acquisition system for the planned U.S.-based Deep Underground Neutrino Experiment.

Agenda: https://indico.cern.ch/event/1167511/