[3Xin4] Poster session 1

[3Xin4-75] Inspection trail with factory data on quantm kernel learning

〇Takao Tomono¹, Satoko Natsubori¹ (1.DI div., Toppan Inc.)

Keywords:quantum kernel, Social implementation, SVM

Machine learning classifiers have been used in medicine, factory inspections, and automated driving. Support Vector Machines (SVMs), one of the classifiers, are useful and have been used in various situations. In particular, kernel methods are very important for nonlinear and unsolvable classification. On the other hand, quantum machine learning has received much attention in recent years, but its specific evaluation has not been done much. In this study, we applied quantum kernel learning to the factory inspection process. As a result, it showed higher performance than classical kernel learning. This time, the image data was preprocessed, binarized, and then subjected to principal component analysis. Although the cumulative contribution rate was 75%, the accuracy was over 97% when performing quantum kernel learning. The accuracy of 93% is also obtained by classical kernel learning. Kernel learning is known to depend on the properties of the dataset, but in the future, we would like to accumulate data on what kind of datasets show the superiority of quantum.