IRIS-HEP Fellow: Mihir Katare
Fellowship dates: May – Aug, 2021
Home Institution: University of Illinois, Urbana-Champaign
Project: Deep Learning Implementations for Sustainable Matrix Element Method Calculations
The Matrix Element Method (MEM) is a powerful statistical analysis technique for experimental and simulated particle physics data. It has several benefits over black-box methods like neural networks, owing to its transparent and interpretable results. The drawback of MEM; however, is the significant amount of computationally intensive calculations involved in its execution, which impedes research that relies on it. This project aims to improve the viability of MEM, by implementing deep learning techniques to accurately and efficiently approximate MEM calculations - providing the much required speedup over the traditional approach, while preserving its interpretability. The implemented model can be used as a good approximation during the exploratory phase of research, and the full ME calculations can be used for the final runs, making the workflow for research involving MEM much more efficient.More information: My project proposal
Mentors:
-
Mark Neubauer (University of Illinois, Urbana-Champaign)
-
Matthew Feickert (University of Illinois, Urbana-Champaign)
- 18 Oct 2021 - "Deep Learning for the Matrix Element Method", Mihir Katare, Recording: Deep Learning for the Matrix Element Method
Current Status
June 2022 - Software Engineer at Amazon Web Services (AWS)
Contact me: