BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//LBNL Physics Division Research Progress Meetings - ECPv6.8.3//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:LBNL Physics Division Research Progress Meetings
X-ORIGINAL-URL:https://rpm.physics.lbl.gov
X-WR-CALDESC:Events for LBNL Physics Division Research Progress Meetings
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:UTC
BEGIN:STANDARD
TZOFFSETFROM:+0000
TZOFFSETTO:+0000
TZNAME:UTC
DTSTART:20260101T000000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=UTC:20260113T160000
DTEND;TZID=UTC:20260113T170000
DTSTAMP:20260414T145427
CREATED:20260105T173222Z
LAST-MODIFIED:20260105T173222Z
UID:3045-1768320000-1768323600@rpm.physics.lbl.gov
SUMMARY:Speaker: Matthias Vigl (Technical University of Munich) - Title: Machine-Learning Scaling Laws for LHC Physics: How Scale Unlocks Breakthrough Gains in Physics Sensitivity
DESCRIPTION:Research Progress Meeting \nDate: January 13\, 2026 \nTime: 4:00- 5:00 pm \nLocation: Sessler Conference Room- 50A-5132 [In-Person and HYBRID]  \nSpeaker: Matthias Vigl (Technical University of Munich) \nTitle: Machine-Learning Scaling Laws for LHC Physics: How Scale Unlocks Breakthrough Gains in Physics Sensitivity \nAbstract: High Energy Physics and deep learning have historically taken different routes to data processing: in collider physics\, performance has been driven by deep\, hand-engineered pipelines that encode decades of domain knowledge\, while modern machine learning has advanced primarily through scale\, leveraging large datasets and increasingly generic model architectures. While machine learning has long been embedded in the HEP analysis pipeline\, the rate of improvement has remained slower than the rapid\, scale-driven progress observed in industry.\nThis talk contrasts physics-driven and scale-driven approaches to data processing and shows how foundation-model principles (scaling laws\, transfer learning\, and end-to-end optimization) can be applied to HEP analyses. Compute-optimal scaling laws are derived for the state-of-the-art ATLAS jet flavor tagger and validated by training models two orders of magnitude beyond previous compute regimes\, yielding predictable performance improvements in line with industry-scale models. When translated into physics sensitivity for flagship ATLAS analyses such as HH(4b) at the High-Luminosity LHC\, these gains correspond to improvements equivalent to multiple years of data taking\, motivating a shift toward large-scale ML model training and deployment within LHC experiments.\n \nJoin Zoom Meeting\nhttps://lbnl.zoom.us/j/95679892182?pwd=RU5xU2dDRFNabnR1U3pQMklkYWFIdz09 \nMeeting ID: 956 7989 2182 \nPasscode: 169037
URL:https://rpm.physics.lbl.gov/event/speaker-matthias-vigl-technical-university-of-munich-title-machine-learning-scaling-laws-for-lhc-physics-how-scale-unlocks-breakthrough-gains-in-physics-sensitivity/
END:VEVENT
END:VCALENDAR