Update, March 2021: Read the official blog for a summary of the project.
This blog serves to document my work during GSoC 2020 on TMVA PyTorch Interface at CERN.
The Toolkit for Multivariate Analysis (TMVA) is a machine learning environment integrated into the ROOT scientific software framework, which has been developed specifically for high-energy physics applications.
TMVA provides functionality for building deep neural networks including fully connected, convolutional and recurrent layers. In addition, TMVA provides interfaces to other deep learning frameworks (scikit-learn, Keras). This project aims to extend these interfaces to one of the most popular, fast and flexible machine learning frameworks, PyTorch. The focus will be on the development of a production-ready TMVA interface to PyTorch, while strictly adhering to my proposal.