Pa Learning Pa Algorithm with Unlabeled Data

We analyze the combination of online learning and semisupervised learning. Online learning is an approach to supervised machine learning, which is effective at processing a large amount of training data. Semi-supervised learning is employed when a large amount of unlabeled data is available. While both setting occurs frequently in real applications of machine learning, thus far, their theoretical aspects have not been fully discussed. Our analysis provides valuable insights into the combination of two approaches.