Convex Formulation for Learning from Positive and Unlabeled Data

We discuss binary classification from only from positive and unlabeled data (PU classification), which is conceivable in various real-world machine learning problems. Since unlabeled data consists of