Machine-learning based particle-flow algorithm in CMS
This work addresses particle reconstruction for high-energy physics experiments like CMS, representing an incremental advancement by applying ML to an existing bottleneck.
The paper tackles the challenge of particle-flow event reconstruction in CMS by developing a machine-learned particle flow (MLPF) algorithm using a transformer model to infer particles from tracks and clusters in a single pass, resulting in improvements in reconstruction metrics as integrated into offline software.
The particle-flow (PF) algorithm provides a global event description by reconstructing final-state particles and is central to event reconstruction in CMS. Recently, end-to-end machine learning (ML) approaches have been proposed to directly optimize physical quantities of interest and to leverage heterogeneous computing architectures. One such approach, machine-learned particle flow (MLPF), uses a transformer model to infer particles directly from tracks and clusters in a single pass. We present recent CMS developments in MLPF, including training datasets, model architecture, reconstruction metrics, and integration with offline reconstruction software.