Using influence functions to investigate generalization capabilities of neural networks. - View it on GitHub