Interpretations are Useful: Penalizing Explanations to Align Neural Networks with Prior Knowledge.

Laura Rieger, Chandan Singh, W. James Murdoch, Bin Yu

Browse the full ICML paper archive.