IRIS-HEP Fellowship Proposal: Deep Learning Implementations for Sustainable Matrix Element Method Calculations

The utilization of multivariate statistical analysis methods in the analysis of experimental and simulated particle physics data has allowed us to gain insight on discoveries of new physics. Methods like neural networks, decision trees, and other machine learning techniques have gained a lot of traction due to their ability to model particle physics phenomena with high accuracy. However, methods like neural networks are effectively black-box methods with limited transparency and interpretability. As such, a lot of effort is usually spent on ensuring the results are generalized, as well as extensively testing whether the network has learnt the underlying physics of the process.