Deep Learning as an opportunity for the future of High Energy Physics
Aula Conversi (Dip. di Fisica - Edificio G. Marconi)
Dip. di Fisica - Edificio G. Marconi
There is a long standing tradition of Machine Learning application in particle physics, dating back to LEP times and going through the key role played at B-factories and Tevatron. Nowadays, deep learning and the technological progresses associated to it (e.g., parallel computing architectures such as FPGAs and GPUs) open up new opportunities for high energy physics beyond the classic particle-identifications and energy-measurement problems. Deep Learning promises to help us solving future challenges (e.g., the large demand of computing power at the High Luminosity LHC) by speeding up computing-heavy tasks like tracking, detector simulation, etc. At the same time, it offers new tools to complement the mainstream strategy to search for physics beyond the standard model. An extensive usage of Deep Learning in central tasks (e.g., on-line decision, detector monitoring, particle reconstruction) will allow to save resources and extend the sensitivity of the next-generation experiments in a more cost-effective way.