Automatic Classification of Tables Using Hybrid Quantum Convolutional Neural Networks

Abstract

An HQCNN was used in this work to show that the outcomes are better than those of a CNN. After presenting the suggested network layout, several hyperparameter combinations were attempted to observe how the outcomes changed. One pretrained CNN experiment (ResNet-18) was selected for the non-quantum portion of the structure after many experiments were conducted with it. It was shown that the HTL improved the classification performance, either for tables or non-tables. The performance is greater when using deep learning with quantum computing, but the computational cost is higher than when using CNNs for testing.