Timezone: »

Explaining machine-learned particle-flow reconstruction
Farouk Mokhtar · Raghav Kansal · Daniel Diaz · Javier Duarte · Maurizio Pierini · jean-roch vlimant

The particle-flow (PF) algorithm is used in general-purpose particle detectors to reconstruct a comprehensive particle-level view of the collision by combining information from different subdetectors. A graph neural network model, known as the MLPF algorithm, has been developed to substitute rule-based PF. However, understanding the model's decision making is not straightforward, especially given the complexity of the set-to-set prediction task, dynamic graph building, and message-passing steps. In this paper, we adapt the layerwise-relevance propagation technique to the MLPF algorithm to gauge the relevant nodes and features for its predictions. Through this we gain insight into the model's decision-making.

Author Information

Farouk Mokhtar (UC San Diego)
Raghav Kansal (UC San Diego)
Daniel Diaz (University of California San Diego)
Javier Duarte (UC San Diego)

I am an Assistant Professor in experimental high energy physics at UC San Diego and a member of the CMS collaboration at CERN. My research interests include measuring the properties and couplings of the Higgs boson and searching for beyond-the-standard-model particles in LHC data. I am interested in developing machine learning algorithms, real-time trigger systems (with applications to embedded devices), and heterogenous computing architectures for the next generation of high energy physics experiments.

Maurizio Pierini (CERN)
jean-roch vlimant (California Institute of Technology)

More from the Same Authors