Learning Broken Symmetries With Approximate Invariance To Better Classify Particle Collision Events
Seth Jacob Nabat, 18, of Winnetka, developed a machine learning program to make sense of the results of particle collisions for his Regeneron Science Talent Search physics project.
Physicists use computer models to study high-energy particle collisions. When these programs expect symmetrical results, it saves computing time and energy, but can lead to measurement errors being doubled. In his project, Seth designed a three-part system to avoid this confusion. The first network is the one that 鈥渒nows鈥� about symmetry and uses it to find the quickest way to the approximate results of collisions. The second, unconstrained network catches camera and measurement errors, and the third network finds patterns in those errors.
He found that, in tests, his combined model was able to navigate the imperfect data without losing the efficiency advantage. His model lets researchers look directly at what 鈥渂reaks鈥� symmetry, a fundamental problem in physics, especially quantum field theory.
The son of Robyn Brook-Nabat and Scott Nabat, Seth attends William Howard Taft Charter High School (Woodland Hills), where he competes in varsity debate as his team鈥檚 co-captain.
Beyond the Project
As a volunteer instructor with the UCLA Math Circle, Seth teaches classrooms of elementary and middle school students college-level mathematics and helps other instructors prepare for their classes.
FUN FACTS: The first time he rode a mule on a family trip, Seth was bucked off when the animal got spooked by a deer carcass.
