Carnegie Mellon University
August 05, 2022

Physics Graduate's Pioneering Work Recognized by CERN Experiment

By Kirsten Heuring

Jocelyn Duffy
  • Associate Dean for Communications, MCS
  • 412-268-9982

What started as a side project by Carnegie Mellon University's Michael Andrews has turned into a pioneering use for machine learning in physics. Andrews, who recently earned his Ph.D. in physics, said members of the Compact Muon Solenoid (CMS) experiment that operates at the Large Hadron Collider at CERN in Switzerland, would use processed data and neural networks to advance research in particle physics.

"Michael is an incredibly good physicist," said Manfred Paulini, professor of physics and Andrews' adviser and a CMS researcher. "Machine learning is a tool, and with every tool, you have to check and calibrate it. Michael knows how to use these tools and appreciate them."

Paulini and other researchers who are part of the CMS experiment previously used processed data and neural networks to advance their research in particle physics. However, a couple of years into his doctoral studies, Andrews said he realized how powerful machine learning could be. He started exploring how it could be applied to his and Paulini’s research.

"I was keen to reassure Manfred that this was just a side project, and it would not distract from my original thesis," Andrews said. "Things matured and showed promise, and at a certain point, we both agreed that it would be more fruitful if we switched thesis topics to something that really sold the strategy on using machine learning on raw sensor data."

Since no one had used raw data and machine learning in this area of particle physics before, Andrews created many of the processes from scratch. He looked into details of disentangling two photons, rays of light, that overlap in the detector previously reconstructed as a single photon with conventional analysis techniques. This process was extremely laborious, and he had to extend his Ph.D. studies. However, thanks to his hard work, CMS now has a new way to analyze data and gain more understanding of high-energy collisions.

For efforts, the CMS collaboration recognized him with the CMS Thesis Award. The award honors three graduating doctoral students who work on the CMS experiment.

"CMS is a very big collaboration. It's very hard to do anything new," Andrews said. "The mindset has changed from when I started to where it is now. When I first started with this, there was a lot of pushback with this idea, but today, that’s changed. We were the first to demonstrate how using raw sensor data could really not just work but really make a big difference."

Andrews wrapped up both his Ph.D. and his postdoctoral studies with Paulini. He now works at Theorem, a startup created by CMU alumnus Hugh Edmundson, where he applies his knowledge of machine learning to new problems in finance. However, he has left a lasting impact on physics research.

"Before Michael, I had ignored machine learning for a very long time," Paulini said. "He was really the driving force of getting into what we can do with respect to machine learning. This is the beauty of being a professor, you can learn together with your students."

Andrews previously won the Guy C. Berry Graduate Research Award from MCS for his work.

— Related Content —