Obtaining Example-Based Explanations from Deep Neural Networks