Tensor Network-Based Quantum-Train with Distributed Ansatz for Scalable Quantum-Classical Models

Published in ICASSP 2025 – IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025

This paper enhances the Quantum-Train (QT) framework by integrating a tensor network model and a distributed circuit ansatz for scalable hybrid quantum-classical learning. Key innovations include:

  • Replacing the traditional multi-layer perceptron (MLP) with a tensor network, improving scalability and interpretability.
  • Leveraging multiple small QPU nodes to support large-scale quantum machine learning tasks.
  • Maintaining low parameter complexity and ensuring quantum-free inference, critical for practical deployment.

Experimental benchmarks demonstrate improved performance and compactness, positioning this framework as a robust solution for scalable quantum-classical model training.

Figure: Quantum-Train pipeline using distributed QNN modules and tensor networks for CNN weight generation.

Recommended citation: Liu, Chen-Yu; Lin, Chu-Hsuan Abraham; & Chen, Kuan-Cheng. (2025). "Tensor Network-Based Quantum-Train with Distributed Ansatz for Scalable Quantum-Classical Models." Proceedings of ICASSP 2025 – IEEE International Conference on Acoustics, Speech and Signal Processing, 1–4.
Download Paper