Identifying Untrustworthy Predictions in Neural Networks by Geometric Gradient Analysis

Open in new window