Using Machine Learning to Emulate Nuclear and Particle Physics Detector Signals
Sydney Ruppert
Undergraduate Researcher
Marketing Major (Kelley School of Business)
Daniel Salvat
Faculty Mentor
Daniel Salvat (College of Arts & Sciences)
Project Description
When a particle interacts with a detector, the detector produces some sort of electrical response, such as a pulse or a frequency signal, which is recorded as a digital waveform. These waveforms are analogous to digitally recorded audio, and machine learning has been widely and successfully applied to both audio recognition and synthetic audio generation. At the same time, there is increasing need to produce synthetic or simulated detector waveforms to analyze along with the real waveforms in order to understand any unforeseen effects or issues with the analysis process. My goal is to take measured digital waveforms from existing detectors and use standard machine learning packages to explore which most accurately reproduce the data and compare with more conventional simulation techniques which are not based upon machine learning. I will work with the student to teach the basic needed skills, provide the necessary scientific context, provide the set of measured waveforms for the different detectors, work with the student to utilize promising machine learning models, and establish benchmarks for reproducibility to set goals for the research throughout the course of the year. This can be done entirely remotely if required.
Technology or Computational Component
The goal is to utilize machine learning to create synthetic data, and is nearly entirely computational. Moreover, I plan to explore with the student machine learning frameworks which enjoy broad usage, thus providing the student with transferable skills which would aide the pursuit of other career paths involving computing and AI.