Pre-training Tensor-Train Networks Facilitates Machine Learning with Variational Quantum Circuits

Variational quantum circuit (VQC) is a promising approach for implementing quantum neural networks on noisy intermediate-scale quantum (NISQ) devices. Recent studies have shown that a tensor-train network (TTN) for VQC, namely TTN-VQC, can improve the representation and generalization powers of VQC. However, the Barren Plateau problem leads to the gradients of the cost function vanishing exponentially small as the number of qubits increases, making it difficult to find the optimal parameters for the VQC. To address this issue, we put forth a new learning approach called Pre+TTN-VQC that builds upon the TTN-VQC architecture by incorporating a pre-trained TTN to alleviate the Barren Plateau problem. The pre-trained TTN allows for efficient fine-tuning of target data, which reduces the depth of the VQC required to achieve good empirical performance and potentially alleviates the training obstacles posed by the Barren Plateau landscape. Furthermore, we highlight the advantages of Pre+TTN-VQC in terms of representation and generalization powers by exploiting the error performance analysis. Moreover, we characterize the optimization performance of Pre+TTN-VQC without the need for the Polyak-Lojasiewicz condition, thereby enhancing the practicality of implementing quantum neural networks on NISQ devices. We conduct experiments on a handwritten digit classification dataset to corroborate our proposed methods and theorems.