Quantum Convolution for Convolutional Neural Networks

Quantum machine learning has garnered a lot of attention recently due to the quick advancement of quantum technologies. For enhancing the performance of classical neural networks, a family of hybrid quantum-classical neural networks made up of both classical and quantum components has received extensive study. A new design of quantum convolutional neural networks (QCNNs), is what we propose in this research. With the help of our technique, the idea of convolution, which is frequently used in contemporary deep learning algorithms, is developed with quantum operations to be used in designing the quantum convolutional neural networks (QCNNs). While lowering the computational cost, the suggested QCNNs are able to capture more context throughout the quantum convolution process. We conduct practical studies on the keras digits dataset to perform image recognition and show that QCNN models generally outperform existing quantum convolutional neural networks (QCNNs) in terms of accuracy and loss computation.