Reproduce “Federated Learning with Quantum Enhanced LSTM for Applications in High Energy Physics” on OpenResearch

About this paper

Abhishek Sawaika, Durga Pritam Suggisetti, Udaya Parampalli, Rajkumar Buyya

Researchers at The University of Melbourne and collaborators developed a federated learning framework that integrates a hybrid quantum-classical Long Short-Term Memory (QLSTM) model for event classification in High Energy Physics. This framework achieved comparable predictive performance (AUC within 1%) to classical deep learning methods while utilizing 100 times less training data and 100 times fewer trainable parameters.

Key insights

• The QLSTM model's hybrid design, integrating VQCs, allows it to effectively capture complex non-linear correlations, leading to enhanced performance over both classical LSTMs and standalone VQCs in HEP data classification.
• Federated learning is a viable paradigm for distributed, privacy-preserving model training in large-scale scientific collaborations like HEP, enabling quantum-enhanced models to learn from decentralized data without direct sharing.
• Quantum-enhanced models like QLSTM demonstrate high representational capabilities, achieving performance on par with large classical deep learning benchmarks using significantly fewer data points and trainable parameters, highlighting potential resource efficiency in the NISQ era.

Summary by alphaXiv.