Quantum semi-supervised kernel learning

Quantum machine learning methods have the potential to facilitate using extremely large datasets. While availability of data for training models is steadily increasing, oftentimes it much easier collect feature vectors obtain corresponding labels. One approaches addressing this issue use semi-supervised learning, which leverages not only labeled samples, but also unlabeled vectors. Here, we present a quantum algorithm kernel support vector machines. The uses recent advances in sample-based Hamiltonian simulation extend existing LS-SVM handle term loss. Through theoretical study algorithm’s computational complexity, show that maintains same speedup as fully-supervised LS-SVM.