Explaining machine-learned particle-flow reconstruction

TL;DR

This paper applies layerwise-relevance propagation to a graph neural network-based particle-flow algorithm, providing interpretability and insights into its decision-making process in particle reconstruction.

Contribution

It introduces a method to interpret GNN-based particle-flow algorithms, enhancing understanding of their complex decision processes.

Findings

Layerwise-relevance propagation successfully identifies relevant nodes and features.

The interpretability method offers insights into the GNN's decision-making.

The approach improves transparency of machine-learned particle reconstruction.

Abstract

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 (GNN) model, known as the machine-learned particle-flow (MLPF) algorithm, has been developed to substitute the rule-based PF algorithm. 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 for GNNs and apply it to the MLPF algorithm to gauge the relevant nodes and features for its predictions. Through this process, we gain insight into the model's decision-making.