Seminar

Speaker: Dr. Spandan Mondal (Brown University)

Title: Machine Learning in Flavour Tagging: Current Advances and Future Directions

Abstract:
Flavour tagging, that is the identification of particle jets arising from heavy-flavour quarks (bottom and charm), is an essential component of several particle physics measurements and searches for physics beyond the Standard Model. The advent of Machine Learning (ML) has revolutionized the field of flavour tagging. From simple fully connected networks to modern transformer-based architectures, flavour tagging algorithms have utilized increasingly complex designs and larger simulated datasets to achieve unprecedented accuracies in heavy-flavour jet identification. ML has also been leveraged for calibrating these algorithms using real collision data, employing techniques such as domain adaptation, adversarial training, and neural network-based optimal transport. The next decade is poised to see the adoption of even newer ML architectures, combined with the use of larger simulated datasets and data-aware training strategies, promising improved sensitivities to rare physics processes at the LHC, HL-LHC, and beyond.