Learning Dynamics of Decision Boundaries without Additional Labeled Data

We propose a method for learning the dynamics of the decision boundary to maintain classification performance without additional labeled data. In various applications, such as spam-mail classificati