Quantum-Train Long Short-Term Memory: Application on Flood Prediction Problem

This study applies the Quantum-Train (QT) framework to enhance Long Short-Term Memory (LSTM) models for flood prediction, a task of growing urgency under climate change. Key features of this approach include:

• Polylogarithmic parameter reduction via quantum machine learning during training, enabling efficient model compression.
• Direct processing of classical data without the need for quantum embedding.
• A fully classical inference stage, removing the requirement for quantum hardware after training.

Originally validated in QHack 2024, the QT-enhanced LSTM delivers a practical and scalable solution for improving flood forecasting accuracy, ultimately supporting better preparedness and disaster response systems.

Recommended citation: Lin, Chu-Hsuan Abraham; Liu, Chen-Yu; & Chen, Kuan-Cheng. (2024). "Quantum-Train Long Short-Term Memory: Application on Flood Prediction Problem." Proceedings of the 2024 IEEE International Conference on Quantum Computing and Engineering (QCE).
Download Paper