Programming Variational Quantum Circuits with Quantum-Train Agent

This paper presents QT-QFWP, a hybrid framework for programming Variational Quantum Circuits (VQCs) using Quantum-Train-driven updates. The model architecture features:

• A quantum-controlled slow programmer that manages classical parameters for a fast-programmer VQC, enabling dynamic and data-adaptive circuit configuration.
• Parameter compression of 70–90% compared to QLSTM and traditional QFWP, significantly improving scalability.
• Superior performance on time-series forecasting tasks like Damped SHM, NARMA5, and simulated gravitational waves.

QT-QFWP offers a viable solution for deploying VQCs on near-term quantum hardware, addressing limitations in qubit count and gate fidelity while maintaining prediction accuracy and runtime efficiency.

Recommended citation: Liu, Chen-Yu; Chen, Samuel Yen-Chi; Chen, Kuan-Cheng; Huang, Wei-Jia; & Chang, Yen-Jui. (2025). "Programming Variational Quantum Circuits with Quantum-Train Agent." Proceedings of the 2025 International Conference on Quantum Communications, Networking, and Computing (QCNC), 544–548.
Download Paper