Project:	High Energy Physics and Machine Learning Projects with the CMS Experiment at the Large Hadron Collider
Disciplines:	Physics, Machine Learning
Mentor:	Harvey Newman, Professor of Physics, (PMA), newman@hep.caltech.edu, Phone: 6263609495
AO Contact:	Raghav Kansal (Co-Mentor), rkansal@caltech.edu
Background:	There are multiple opportunities available at the intersection of experimental particle physics with CMS at the CERN Large Hadron Collider (LHC), and using artificial intelligence (AI) / machine learning (ML) for improving analysis and computing techniques. These opportunities are available for both SURF and term-time research all year round, for Caltech students only.
Description:	Example Projects Project 1: This project aims to boost the search for high energy pairs of Higgs bosons using state-of-the-art deep learning techniques for identifying Higgs decays to tau leptons. Searching for Higgs boson pairs allows us to understand the dynamics of the very early universe, through measurement of the Higgs self-coupling and thereby the shape of the Higgs potential. This is why measuring HIggs pair production (HH) is a flagship physics target of the LHC in the coming decades. This project explores a novel data analysis strategy to do so, looking for highly Lorentz-boosted Higgs bosons decays to two bottom quarks (bb) and two tau leptons (ττ). H→ττ is a very rich and complicated decay, and we aim to hep.explore ML techniques such as graph and transformer networks to improve its identification and reconstruction. Project 2: Simulations are critical in particle physics to understanding what we expect to observe in our experiments. However, traditional methods can be computationally intensive and likely will not be able to scale to the exponentially increasing data volume we expect in the coming decades. In this project, we aim to use generative AI techniques, such as diffusion and flow-matching models that have shown great promise in computer vision, to simulate particle collisions.
References:	On Higgs pair production using machine learning: 1. CMS paper on HH→4b https://arxiv.org/abs/2205.06667 2. Recorded talk on CMS HH→4b https://indico.fnal.gov/event/55499/ 3. Transformers for “jet tagging” https://arxiv.org/abs/2202.03772 On machine learning for CMS simulations: 1. Graph-based generative adversarial networks for jet simulations https://arxiv.org/abs/2106.11535 2. Tutorials on diffusion models and flow matching: https://cvpr2022-tutorial-diffusion-models.github.io/, https://jmtomczak.github.io/blog/18/18_fm.html
Student Requirements:	Introductory physics courses and coding experience. Familiarity with high energy physics, machine learning and/or previous project experience is a plus.
Programs:	This AO can be done under the following programs:   Program    Available To        SURF    Caltech students only  Click on a program name for program info and application requirements.