Learning to Explain: An Information-Theoretic Perspective on Model Interpretation