#AAAI2026 Invited Talk: Machine Learning for Particle Physics

Why It Matters

By breaking the helical‑track constraint, machine learning could expose previously hidden phenomena, accelerating breakthroughs in fundamental physics and expanding AI’s role in scientific discovery.

Key Takeaways

• LHC collisions operate at 13 TeV, probing extreme energies
• ML has evolved from shallow networks to deep learning
• Graph neural networks separate hit clustering from trajectory fitting
• Non‑helical track detection could reveal particles like quirks
• Whiteson links particle physics research with public science outreach

Pulse Analysis

Machine learning has become a cornerstone of particle physics, evolving from early dimensionality‑reduction tools in the 1990s to the deep neural networks that helped confirm the Higgs boson in 2012. Today, AI not only classifies collision events but also generates realistic simulated data, enabling researchers to explore parameter spaces that would be infeasible with traditional Monte Carlo methods. This convergence of high‑energy experiments and advanced analytics has created a feedback loop: better models improve detector design, and richer data fuels more sophisticated algorithms.

A persistent bottleneck lies in the way particle tracks are reconstructed. Conventional pipelines assume charged particles follow helical paths within the LHC’s magnetic field, an assumption that simplifies computation but discards any non‑standard trajectories. Whiteson’s work leverages graph neural networks to decouple hit clustering from trajectory fitting, mapping detector signals into a latent space where hits from the same particle naturally group together. This approach can accommodate arbitrary smooth paths, opening the door to identify exotic signatures such as oscillating “quirk” particles that defy the helix model.

The broader implications extend beyond academia. Demonstrating that AI can relax long‑standing physics constraints signals a new paradigm where data‑driven discovery supersedes handcrafted heuristics. Industries focused on complex pattern recognition—ranging from autonomous navigation to genomics—can draw lessons from these graph‑based techniques. Moreover, the public‑facing aspect of Whiteson’s career, including books, comics, and podcasts, helps demystify cutting‑edge science, fostering a talent pipeline that will sustain interdisciplinary innovation at the intersection of AI and fundamental research.