The Transformer and its uses in High Energy Physics

par Corentin Allaire (IJCLab, Université Paris-Saclay, CNRS/IN2P3)

jeudi 22 oct. 2026, 14:00 → 15:00 Europe/Paris

Amphi G.Besse (Subatech IMT Atlantique)

The transformer network, proposed by Google scientists in 2026 in the « Attention Is All You Need », has become a game-changer in the field of Deep learning. Originally developed for natural language translation, they became the cornerstone of all modern Large Language Models. While those models are extremely good at handling language-based input, they can also be quite performant at handling physics data. In this seminar, after a small reminder on the basics of deep learning, we will be introducing the Transformer, how it works, how one can build/train one from scratch and what tasks it is well-suited for. All this while, taking, for example, the field of High Energy Physics. A domain well known for its large volume of data and pioneering use of Deep Learning technology.