Unmasking the Higgs potential: next-generation machine learning for tracking, flavour tagging, and discovery

Ziel

The discovery of the Higgs boson was a landmark achievement in particle physics, but many fundamental questions remain unanswered, especially how the Higgs boson interacts with itself. Measuring this self-interaction will reveal how the Higgs field shaped the early Universe and why matter exists in abundance over antimatter. Deviations from expected values could shed light on the origin of dark matter and cosmic inflation. This project focuses on the most sensitive way to study this: the simultaneous production of two Higgs bosons (“di-Higgs” production).

Observing di-Higgs production is extremely challenging due to its rarity and the complexity of its experimental signatures. The High-Luminosity Large Hadron Collider (HL-LHC) will soon deliver the only dataset capable of probing the Higgs self-coupling for the foreseeable future, yet current reconstruction methods may fall short. At the same time, machine learning has entered a new era: transformer and segmentation models, already revolutionising natural language processing and medical imaging, are now ready to transform collider physics. The convergence of these developments creates a unique opportunity, which I am ideally positioned to seize.

I have pioneered the use of graph neural networks and transformers in ATLAS, spearheaded flavour tagging innovations, and led Higgs analyses. This rare combination of expertise, spanning reconstruction, machine learning, and Higgs physics, equips me to lead a conceptual leap.

UNMASK will re-imagine the core reconstruction tasks of charged particle tracking and b-jet identification as holistic segmentation problems, inspired by state-of-the-art architectures from computer vision such as MaskFormer. By overcoming the limits of today’s modular algorithms, it will unlock unprecedented sensitivity to di-Higgs production, while delivering broadly applicable tools for the HL-LHC and future colliders, establishing a new paradigm for reconstruction in particle physics.

Wissenschaftliches Gebiet (EuroSciVoc)

Dieses Projekt wurde noch nicht bei EuroSciVoc klassifiziert.
Schlagen Sie die Wissenschaftsbereiche vor, die Ihrer Einschätzung nach besonders relevant sind, und helfen Sie uns, unseren Klassifizierungsdienst zu verbessern.