Training Data Influence Analysis and Estimation: A Survey

Abstract

Good models require good training data. For overparameterized deep models, the causal relationship between training data and model predictions is increasingly opaque and poorly understood. Influence analysis partially demystifies training's underlying interactions by quantifying the amount each training instance alters the final model. Measuring the training data's influence exactly can be provably hard in the worst case; this has led to the development and use of influence estimators, which only approximate the true influence. This paper provides the first comprehensive survey of training data influence analysis and estimation. We begin by formalizing the various, and in places orthogonal, definitions of training data influence. We then organize state-of-the-art influence analysis methods into a taxonomy; we describe each of these methods in detail and compare their underlying assumptions, asymptotic complexities, and overall strengths and weaknesses. Finally, we propose future research directions to make influence analysis more useful in practice as well as more theoretically and empirically sound. A curated, up-to-date list of resources related to influence analysis is available at https://github.com/ZaydH/influence_analysis_papers.

Published in Springer journal "Machine Learning"

References in corpus (19)

• Kernel methods in machine learning
• Understanding deep learning requires rethinking generalization
• A Survey on the Explainability of Supervised Machine Learning
• Poisoning Attacks against Support Vector Machines
• Sketching as a Tool for Numerical Linear Algebra
• Data Shapley: Equitable Valuation of Data for Machine Learning
• What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation
• A Survey of Machine Unlearning
• From Shapley Values to Generalized Additive Models and back
• Influence-guided Data Augmentation for Neural Tensor Completion
• Adapting and Evaluating Influence-Estimation Methods for Gradient-Boosted Decision Trees
• RIM: Reliable Influence-based Active Learning on Graphs
• Data Banzhaf: A Robust Data Valuation Framework for Machine Learning
• TracInAD: Measuring Influence for Anomaly Detection
• Run-Off Election: Improved Provable Defense against Data Poisoning Attacks
• On Collective Robustness of Bagging Against Data Poisoning
• Provable Robustness Against a Union of ${\ell }_0$ Adversarial Attacks
• Measuring the Effect of Training Data on Deep Learning Predictions via Randomized Experiments
• Representer Point Selection for Explaining Regularized High-dimensional Models

Cited by in corpus (4)

• Creative Writers' Attitudes on Writing as Training Data for Large Language Models
• Assessing the Efficacy of Deep Learning Approaches for Facial Expression Recognition in Individuals with Intellectual Disabilities
• Leveraging Influence Functions for Resampling Data in Physics-Informed Neural Networks
• SwarmSearch: Decentralized Search Engine with Self-Funding Economy