fairfolio
Saved Projects

Achieving Improved Accuracy Model for Jet Energy Measurements in the Large Hadron Collider (LHC) Using Machine Learning

CSEF · 2018 Physics & Astronomy Honorable_mention Award

Milan Ganai · Lynbrook High School, San Jose, Santa Clara County

Overview

Objectives/Goals At LHC, the largest particle accelerator, the discovery of Higgs boson and confirmation of Higgs fields rattled the world in 2013, sealing a major unverified gap in SM. Energy measurements of jets produced by hadronization of quarks and gluons help to search for important rare physics processes beyond the Standard Model (SM). However, the pileup interactions cause inaccuracies in the measurements of jet mass and transverse momentum. I use machine learning (ML) to achieve improved accuracy in jet energy measurements by investigating more variables in the regression. My hypothesis is that ML will provide more accurate, robust, scalable and faster pileup mitigation than previous widely-used approaches at LHC. Methods/Materials The most used method at LHC today is area-based pileup mitigation, where a term proportional to jet area is subtracted from the measured jet transverse momentum and mass. Although it is an effective approach, it is inadequate in achieving the needed accuracy for next generation colliders due to its simplistic modeling. I propose a novel pileup mitigation methodology for jet energy measurements using TensorFlow-based machine learning. I investigated 11 most influencing parameters having high correlation with jet measurements and included them in regression training. The training set comprises ~350K full jets (with pileup) and hard jets (without pileup), generated using PYTHIA+FastJet software with varying center-of-mass energies and pileup contaminations. Results The error in transverse momentum using my approach was much smaller (mean = 0.53 and variance = 4.88) compared to the area-based approach (mean = -1.42, variance = 29.44). Similar results were obtained for jet mass measurements. Further, my approach is very robust, and performs better over a wider range of transverse momentum, ranging from 7GeV - 300 GeV. The proposed flow is also orders-of-magnitude faster than the area-based. Conclusions/Discussion I created and used a novel machine learning methodology to build a robust, scalable solution with better accuracy in predicting jet transverse and mass, thereby confirming and validating my hypothesis. Superiority of Deep Neural Network over linear regressors also shows that linear model used in previous area-based approach is inadequate and has limitation in achieving accuracy. Overall, my approach shows a great potential for exploring in next gen SLHC.

Summary statement

I create and use a novel pileup mitigation methodology to improve accuracy in jet energy measurements by studying and using many variables in machine learning and achieve more accurate and faster solution than widely-used approaches at LHC

Help received

I came up with the idea for the project, and developed the methodology and software flow, and analyzed the results. I presented it to my high school science teacher Mr. Leung, who encouraged me to pursue the project.

Awards (1)

  • Honorable Mention

Competition history

  • CSEF 2018 Physics & Astronomy · Entry S1807

Resources

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

Source: California Science & Engineering Fair public projects

Save projects to your library

Sign in with Google to keep track of projects you find interesting, organized into folders. An account also raises your daily allowance for “Has this been done?”, and lets you create a key for the MCP server with a much higher limit than anonymous use. Browsing stays public.

Continue with Google