PyTorch Implementation of Influence Function

Influence function for neural networks is proposed in the ICML2017 best paper (Wei Koh & Liang, 2017). However, to the best of my knowledge, there is no generic PyTorch implementation with reliable test codes. Based on some existing implementations, I’m developing reliable Pytorch implementation of influence function.

• https://github.com/ryokamoi/pytorch_influence_functions
• https://github.com/ryokamoi/test_pytorch_influence_functions

My repositories are forks of the following great work.

• https://github.com/nimarb/pytorch_influence_functions
• https://github.com/dedeswim/pytorch_influence_functions
• https://github.com/nayopu/influence_function_with_lissa

References

1. Wei Koh, P., & Liang, P. (2017). Understanding Black-box Predictions via Influence Functions. In International Conference on Machine Learning (ICML).