Particle flow algorithm with Transformer-based algorithm for highly-granular calorimeters

Jul 28, 2025, 5:25 PM
1m
Barus and Holley, Hazeltine Commons (Brown University)

Barus and Holley, Hazeltine Commons

Brown University

184 Hope St., Providence, RI 02912

Speaker

Tatsuki Murata (University of Tokyo)

Description

Particle flow (PFlow) is one of the key detector concepts of future Higgs factories. It went back to more than 15 years ago that a proposal of highly-granular calorimeters with PFlow algorithm giving almost twice better jet energy resolution than traditional calorimetry was published.
The expected performance has not significantly changed since then while the hardware studies have been greatly improved.
We aim to significantly change the situation at the first time in 15 years with a light of recent DNN improvements. After the first trial made with GNN-based algorithm, we are trying to replace it with more recent Transformer-based algorithm. First implementation already gives promising insights. In this poster we will show details of the algorithm, comparison with existing algorithms and future prospects.

Authors

Mr Stefan Barbu (Ecole Polytechnique / The University of Tokyo) Taikan Suehara (ICEPP, The University of Tokyo) Tatsuki Murata (University of Tokyo) Dr Tomohiko Tanabe (MI-6, Ltd.)

Presentation materials