Variational quantum one-class classifier

Abstract One-class classification (OCC) is a fundamental problem in pattern recognition with wide range of applications. This work presents semi-supervised quantum machine learning algorithm for such problem, which we call variational one-class classifier (VQOCC). The suitable noisy intermediate-scale computing because the VQOCC trains fully-parameterized autoencoder normal dataset and does not require decoding. performance compared that support vector (OC-SVM), kernel principal component analysis (PCA), deep convolutional (DCAE) using handwritten digit Fashion-MNIST datasets. numerical experiment examined various structures by varying data encoding, number parameterized circuit layers, size latent feature space. benchmark shows comparable to OC-SVM PCA, although model parameters grows only logarithmically size. outperformed DCAE most cases under similar training conditions. Therefore, our constitutes an extremely compact effective OCC.