Novel machine learning applications at the LHC

Particle physicists have a long history of developing and applying machine learning (ML) techniques. From early applications of neural networks to charged particle tracking in the 1980s [1] to the Higgs boson discovery in 2012, in which boosted decision trees improved the sensitivity to the decay mode [2], ML has changed the way particle physicists conduct searches and measurements. It is an essential and versatile tool that we use to improve existing approaches, and it enables fundamentally new approaches. In recent years, the subfield of ML in particle physics has grown exponentially in the number of publications and expanded to cover a wide variety of topics and use cases, as indexed by the HEP ML Living Review [3]. In these proceedings, we present selected recent results that highlight how LHC experiments are applying novel ML techniques. In particular, we briefly describe the ML techniques and results for improved classification, faster simulation, unfolding, and anomaly detection.