Introduction to Heterogeneous Knowledge Transfer for Quantum Convolutional Neural Networks

The convolutional neural network (CNN) is widely utilized in computer vision due to its ability to effectively harness correlation information within data. However, when the data's dimensionality or the model's complexity becomes excessively large, CNN faces significant challenges in maintaining efficient learning. The quantum convolutional neural network (QCNN) offers an innovative solution by leveraging quantum computing environments, to address problems conventionally solved by CNN or enhance the performance of existing learning models. This paper proposes a model of the CNN structure for quantum computing and introduces a method for enhancing model performance by integrating heterogeneous knowledge transfer (HKT) for QCNN with the CNN model used in conventional computer vision tasks. Furthermore, this paper examines the feasibility of the QCNN-HKT model in comparison to CNN and QCNN by conducting training experiments on the KITTI dataset with quantum computing simulation libraries Torchquantum.