FPGA-based Firmware Implementation of MET Machine Learning Algorithm for the CMS Phase-2 Level-1 Correlator Trigger

This report summarizes my work as a CERN 2021 summer student, specifically my contributions to a machine learning MET algorithm which potentially will be implemented in the CMS’s level-1 correlator trigger as part of the phase-2 upgrades. I discuss several background topics and the general status of the project after compressing the model and developing a data generator. I show how we quantify the model’s predictive performance and share data from the most recently trained model. From this, I state several conclusions and ultimately determine that with more training data, L1METML will further outperform PUPPI, the MET algorithm currently in use at CMS.