Speaker
Description
Search for physics beyond the Standard Model has been a long-standing subject at the LHC. The absence of such signatures indicates that new physics may elude standard triggers; conventional triggers at the ATLAS experiment are constructed by setting thresholds on variables such as the particle momentum, targeting event topologies exclusive to specific models. Anomaly detection, a form of unsupervised machine learning, enables searches for signatures which deviate from the Standard Model without relying on particular model assumptions. We present the first anomaly detection trigger at ATLAS, newly developed and integrated for data-taking in LHC Run 3. In addition to its design and expected performance, we discuss its commissioning, validation, and operational robustness, along with some look in the newly collected data. The first anomaly detection trigger in ATLAS marks a milestone for machine learning-based, next-generation triggers and model-agnostic searches for new physics.
