hardware testing & honors thesis

overview

One of the requirements for honors physics students at UofM is to complete a senior honors thesis. When I returned to Michigan after my research projects at CERN and with the UZH CMS group, I was interested in continuing to pursue research at the interface of machine learning and particle physics.

I was very lucky to meet University of Michigan professor Christine Aidala, who kindly agreed to supervise my honors thesis. I was also given the chance to contribute to some of the research projects going on within her group.

sPHENIX hardware testing

One of the ongling projects in the group was an effort to test a number of Silicon Photomultipliers (SiPMs) for the sPHENIX detector at the RHIC. Though this process was largely automated, it also required a human operator to take shifts testing. I took a weekly shift assignment for the group.

While the work of overseeing the hardware testing process was not too intensive, I spent a lot of time in the lab discussing research projects with the graduate students in the group, which helped inform my decision to pursue PHD programs.

ML related things

A good amount of people in the group were interested in implementing machine learning analyses, so I had quite a few very interesting conversations about different algorithms and their possible usefulness for physics.

sPHENIX particle identification

This was a project being worked on by Dr. Sookhyun Lee & undergraduate Alan Tondryk. The goal was to determine whether machine learning might be used to improve the particle identification software for the upcoming sPHENIX experiment.

Since I had some experience in implementing machine learning for physics data analysis, I was interested in the project and met with the group weekly. My main contribution was to help with some of the syntactic quirks of implementing ML in Python, suggest algorithms I had read of that they might be able to find success with, and help debug ML-related matters. The results were written up and can be found on Alan’s research website.

SPS ML tutorial

I was asked by the Society of Physics Students (SPS) to host a workshop on machine learning for physics, and did so in late January 2020. The materials can be found here.

This was a two-day workshop, and I opted to make it directly interactive by using Google Colab. This meant that students could see and run code in their own Python3 environment, without the need to mess with any environmental factors.

The workshop as a whole covered

• an introduction to Jupyter notebooks & configuring Python environments by running bash commands
• an overview of data processing and loading resources
• a series of advanced visualization techniques
• an overview of ML and its uses (slide format)
• a walkthrough of building/training a simple neural network
• a small model zoo (and a particle physics example!)

It was generally well received, and generated a lot of interesting discussions afterwards!