Hybrid Quantum‐Classical Inception Neural Network for Image Classification

Abstract The advantages of quantum neural networks in image classification have been demonstrated in recent studies, but as the number of network layers increases, the model becomes difficult to train and performs poorly. In addition, research on selecting appropriate quantum circuits for the models is still scarce. Therefore, in order to solve these problems, in this paper, based on parameterized quantum circuits, a hybrid quantum‐classical Inception neural network model for image classification is proposed. This model integrates three parallel feature extraction layers: a quantum layer, a classical layer and a hybrid quantum‐classical layer, which can fully utilize the advantages of quantum high‐dimensional computation and classical large‐scale processing. In addition, the interplay between the expressibility and entanglement of quantum circuits and the model's performance is explored, offering insights for future research. Experimental results on the CIFAR‐10 dataset demonstrate that the model achieves significantly higher classification accuracy compared to several existing approaches.