CURF Introduction: Machine learning in particle physics

Hello! My name is Aleksa, I am a sophomore electrical engineering student, and my project is about using machine learning to create a classifier that can make decisions on a nanosecond scale. My interest in this program began in an Honors Physics class during my freshman year when my professor and current research advisor, Dr. Tae Min Hong, discussed his research during class. His research interested me because the project’s focus on physics and programming is providing me with invaluable experience in my field of study, and I was motivated by the widespread application of machine learning. The primary application for my project is to support data collection at the Large Hadron Collider, the world’s largest and most powerful particle accelerator; however, our machine learning project can be applied to numerous problems outside of physics as well.

The problem guiding our project is that collisions in particle physics experiments occur at an exceedingly high frequency, generating a larger volume of data than the data collector can support; for example, the Large Hadron Collider collides 40 million bunches of protons per second, each of which generates 1 MB of data. As a result, the data must be efficiently filtered through a process called triggering so that only research-relevant events, which occur once every few thousand collisions, are saved. 

One solution to this problem is the Hong group’s fwX, which implements a fast machine learning method called boosted decision trees (BDTs) on configurable electronics called field programmable gate arrays (FPGAs). This implementation can effectively filter data on a scale of nanoseconds for less complicated physics problems with a small number of variables. However, the binning algorithm that fwX uses to categorize the events is currently incapable of handling more complex problems that would require a larger maximum tree depth, which is the maximum number of decisions that the tree would have to make while evaluating an event. As a result, my goal for the Spring semester is to implement a new algorithm that can help fwX perform classification for more complicated problems in physics.

Receiving the CURF will be pivotal for my future academic and career goals because of the real-world experience I will obtain utilizing my studies in one of the world’s most important physics applications. Regardless of whether I decide to pursue a higher degree or a career after I complete my undergraduate degree, the funding and research opportunity provided to me by the CURF will grant me invaluable experiential knowledge that will benefit me regardless of my future path.

Finally, one unique fact about me is that I was born in Serbia and became an American citizen seven months ago! I am very grateful for the CURF award, which has provided me with enough money to upgrade to a more powerful computer that can handle the research group’s technology, and I look forward to working with Dr. Hong and his team this semester!

Leave a Reply