Active Teaching for Inductive Learners

We propose and study a new intelligent teaching paradigm called active teaching in this paper. In contrast to active learning, we assume that the learner can only passively conduct inductive learning from the given examples, but the teacher (oracle) can actively provide “good” examples to the learner, in order to speed up the teaching (learning) process. We establish the framework with four specific paradigms of active teaching, and develop the corresponding teaching strategies. Empirical study shows that, the proposed teaching strategies can indeed outperform the traditional active learning and the basic random sampling strategies, thus making the teaching (learning) process more efficient.