Speaker
Description
Graph Neural Networks (GNNs) are increasingly used for particle tracking in High Energy Physics (HEP), as they provide a natural framework for modeling the relational structure of detector hits. In parallel, recent developments in quantum computing have motivated the exploration of quantum machine learning techniques, which may offer enhanced expressive power through quantum superposition and entanglement. However, the impact of quantum-enhanced models on realistic HEP tracking tasks remains largely unexplored.
In this contribution, we study a hybrid GNN architecture that incorporates variational quantum circuits for edge classification in particle tracking. Quantum layers are embedded within a multilayer perceptron acting on graph edges, and the number of qubits and circuit depth are systematically varied to assess scalability and model expressiveness within the constraints of near-term quantum devices. The hybrid approach is compared to a purely classical GNN with an otherwise identical architecture. Model performance is assessed using standard tracking metrics, including accuracy, F1 score, AUC–ROC, as well as computational cost in terms of training time, inference time, and parameter count.
This study provides a systematic comparison of classical and quantum-enhanced GNN architectures for particle tracking, and establishes a benchmark for future investigations of quantum machine learning methods in HEP reconstruction workflows.
| I read the instructions above | Yes |
|---|