paper-with-me

홈 › Papers

Understanding Black-box Predictions via Influence Functions

2017-03-14 · ICML 2017 8 · Pang Wei Koh, Percy Liang

How can we explain the predictions of a black-box model? In this paper, we use influence functions -- a classic technique from robust statistics -- to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, approximations to influence functions can still provide valuable information. On linear models and convolutional neural networks, we demonstrate that influence functions are useful for multiple purposes: understanding model behavior, debugging models, detecting dataset errors, and even creating visually-indistinguishable training-set attacks.

📄 PDF Abstract BibTeX arXiv:1703.04730

Code (19)

https://worksheets.codalab.org/worksheets/0x2b314dc3536b482dbba02783a24719fd 공식 구현
4pygmalion/Federated_learning-filtering-non-influence-data tf
AnonymizedAuthor663/NNIF_adv_defense tf
ShinKyuY/Understanding-Black-box-Predictions-via-Influence-Functions-tutorial-MNIST
ShinKyuY/Understanding-Black-box-Predictions-via-Influence-Functions-tutorial-MNIST-7-vs-1-Classification
Shmoo137/Interpretable-Phase-Classification pytorch
TooTouch/WhiteBox-Part2 tf
aai-institute/pyDVL pytorch
alstonlo/torch-influence pytorch
bsharchilev/influence_boosting tf
darkonhub/darkon tf
giladcohen/NNIF_adv_defense tf
goldin2008/modeling-project
kohpangwei/influence-release tf
nimarb/pytorch_influence_functions pytorch
pomonam/kronfluence pytorch
qmkakaxi/FederatedLearning pytorch
ryokamoi/pytorch_influence_functions pytorch
sillwood/voicemos pytorch

Similar Papers 제목 키워드 기반

On Second-Order Group Influence Functions for Black-Box Predictions

2019-11-01 · ICML 2020 1 · Samyadeep Basu, Xuchen You, Soheil Feizi

With the rapid adoption of machine learning systems in sensitive applications, there is an increasing need to make black-box models explainable. Often we want to identify an influential group of training samples in a par…

BIG-bench Machine Learning

CLIF: Concept-Level Influence Functions for Transparent Bottleneck Models

2026-05-19 · Yike Sun, Mingkun Xu, Mu You, Zhongzhi He 외 arxiv

In recent years, the black-box nature of deep learning models has limited their application in high-stakes domains such as medical diagnosis and finance, where interpretability is essential. To address this, we propose a…

Medical Diagnosis

Deeper Understanding of Black-box Predictions via Generalized Influence Functions

2023-12-09 · Hyeonsu Lyu, Jonggyu Jang, Sehyun Ryu, Hyun Jong Yang

Influence functions (IFs) elucidate how training data changes model behavior. However, the increasing size and non-convexity in large-scale models make IFs inaccurate. We suspect that the fragility comes from the first-o…

Influence ApproximationPhilosophy

Revisiting Data Attribution for Influence Functions

2025-08-10 · Hongbo Zhu, Angelo Cangelosi arxiv

The goal of data attribution is to trace the model's predictions through the learning algorithm and back to its training data. thereby identifying the most influential training samples and understanding how the model's b…

Explaining Black Box Predictions and Unveiling Data Artifacts through Influence Functions

2020-05-14 · ACL 2020 6 · Xiaochuang Han, Byron C. Wallace, Yulia Tsvetkov

Modern deep learning models for NLP are notoriously opaque. This has motivated the development of methods for interpreting such models, e.g., via gradient-based saliency maps or the visualization of attention weights. Su…

Natural Language Inference