Physics at the Large Hadron Collider: A New Window on Matter, Spacetime and the Universe II
summary
Tracking currently uses seeding plus a Kalman filter with combinatorics that explode as pileup rises, threatening CPU budgets and motivating new pattern‑recognition approaches and a community tracking challenge. Full GEANT simulations cost billions of CPU hours, and generative adversarial networks produce calorimeter showers 1.8 s → ~12 μs on GPUs. Convolutional nets on NOvA-like images give 40% better electron efficiency at fixed background; jet and photon classifiers cut backgrounds by factors up to ~150 and already improve heavy‑flavor tagging and physics limits.
AI-written summary · about
chapters (13)
- 0:00Start
- 5:39Calorimeter Reconstruction
- 9:57Nova Experiment
- 13:49Deep Topology Classifiers
- 14:37Abstract Image Classification
- 19:48Particle Sequence Classifier
- 21:18Results
- 28:29The Generative Adversarial Network
- 31:16Tracking Machine Learning Challenge
- 35:00The Higgs Machine Learning Challenge
- 39:59High-Energy Physics Tracking
- 46:58Limit on the Production Rate
- 51:54Takeaway Messages
related talks
20mAI in the physical world: from observation to discovery
16mGraph Network in High Energy Physics
20mAI-Driven Real-Time Ball Tracking for Live Sports Streaming
59mSimulating the Universe Using Modern C++
43mBlack holes and gravitational waves are discovered (also) with Python!
44mClassical algorithm for simulating experimental Gaussian boson sampling
37mTraining and Optimisation of Large Transformer Models: An ATLAS and CERN Use Case - Ricardo Rocha, CERN & Maxence Draguet, University of Oxford
13mSimulating the Physical World at Scale: Humanoid Locomotion with Ray