Deep learning and high-performance computing methods are becoming key in analysing some elementary particles of the Standard Model: the neutrinos. Current neutrino experiments are leveraging these techniques, which in combination have exhibited to outperform standard tools in several domains, such as identifying neutrino interactions or reconstructing particle energies. In this seminar, he will show various deep-learning algorithms used in the context of the T2K and DUNE experiments, two cutting-edge international neutrino experiments. He will present how transformers, recurrent-neural networks, and sparse convolutional networks are used for tasks including particle tracking, particle identification and electron/gamma separation. All of these methods report promising results and are already integrated (or planned to be) into the reconstruction chain of the experiments.

Lastly, he will present a detailed study on the performance of some of the models above on hardware accelerator boards.

Saúl Alonso-Monsalve is a senior software engineer in the Rubbia Group of the Department of Physics at ETH Zurich (Switzerland), where he leads the efforts in artificial intelligence applied to neutrino physics at the international level. He obtained his PhD in Computer Science from Universidad Carlos III de Madrid (Spain) in 2021 for his research at CERN (European Organization for Nuclear Research, in Switzerland).

He was awarded two prestigious Spanish national awards for the quality of his doctoral thesis: the Spanish Computer Science Society (SCIE) - BBVA Foundation Research Award, for the innovative and relevant doctoral research of young computer scientists; and the Enrique Fuentes Quintana - Funcas Award, for the best thesis at the national level in the category of Engineering, Mathematics, Physics.