Quantum anomaly detection for collider physics.

Area of Science:

• High Energy Physics
• Quantum Computing
• Machine Learning

Background:

• Anomaly detection is crucial for discovering new physics at the Large Hadron Collider (LHC).
• Semi-supervised learning approaches are suitable for analyzing complex datasets with limited labeled examples.
• Quantum Machine Learning (QML) is a potential candidate for enhancing data analysis in high-energy physics.

Purpose of the Study:

• To investigate the efficacy of Quantum Machine Learning (QML) for anomaly detection in a specific high-energy physics context.
• To compare the performance of QML algorithms against Classical Machine Learning (CML) benchmarks in a semi-supervised setting.
• To assess the potential for quantum advantage in classification tasks with small training datasets.

Main Methods:

• A semi-supervised learning approach was applied to simulated four-lepton final state data from the LHC.
• Standard QML algorithms were implemented and evaluated for anomaly detection.
• CML algorithms were used as benchmarks for performance comparison.
• Injected anomalous events were used to test the models' ability to identify deviations from background processes.

Main Results:

• Classical Machine Learning (CML) benchmarks demonstrated superior performance compared to standard QML algorithms.
• CML models were capable of automatically identifying anomalous events within background-only datasets.
• The study explored a scenario (small training datasets) where quantum advantage is theoretically predicted.

Conclusions:

• Current CML methods are effective and outperform QML for this specific anomaly detection task at the LHC.
• The potential for quantum advantage in this regime requires further investigation with more advanced QML algorithms or different problem formulations.
• The findings highlight the importance of robust CML benchmarks in evaluating emerging quantum computing applications.