If Influence Functions are the Answer, Then What is the Question?

Juhan Bae, Nathan Hoyen Ng, Alston Lo, Marzyeh Ghassemi, Roger Baker Grosse

Keywords: Influence Function, Interpretability

Abstract: Influence functions efficiently estimate the effect of removing a single training data point on a model's learned parameters. While influence estimates align well with leave-one-out retraining for linear models, recent works have shown this alignment is often poor in neural networks. In this work, we investigate the specific factors that cause this discrepancy by decomposing it into five separate terms. We study the contributions of each term on a variety of architectures and datasets and how they vary with factors such as network width and training time. While practical influence function estimates may be a poor match to leave-one-out retraining for nonlinear networks, we show that they are often a good approximation to a different object we term the proximal Bregman response function (PBRF). Since the PBRF can still be used to answer many of the questions motivating influence functions, such as identifying influential or mislabeled examples, our results suggest that current algorithms for influence function estimation give more informative results than previous error analyses would suggest.

TL;DR: Influence functions align poorly with leave-one-out retraining because they are actually answering a different question we call the proximal bregman response function.

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2209.05364/code)