Hybrid quantum convolutional neural network for multi-channel image classification

As the size and resolution of image data increase, classical machine learning algorithms will face greater computational challenges. The superposition and entanglement properties of quantum computing provide a new paradigm to break through this bottleneck. Therefore, some researchers have combined quantum computing with classical neural networks, and recent work has demonstrated advantages on noisy intermediate-scale quantum (NISQ) devices using hybrid quantum–classical architectures, which have been effective in the classification task of simple images. However, based on restricted quantum resources, some recently proposed quantum convolutional neural network models lack the ability to efficiently process multi-channel images. To address this issue, in this work, a hybrid quantum–classical convolutional neural network architecture is designed for multi-channel image classification, incorporating the Ising coupling gates, which has recently achieved success in gray-scale image classification. Experiments show that the model exhibits higher predictive performance and stability in the CIFAR-10 image classification task.