On the Usage of Gaussian Process for Efficient Data Valuation

2025/06/04 by Clément Bénesse, Patrick Mesana, Bénesse, Clément +5
Computer Science · Decision Sciences · #Bayesian Modeling and Causal Inference #FOS: Computer and information sciences #Forecasting Techniques and Applications #Gaussian Processes and Bayesian Inference #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.2506.04026

openalex publication_date 2025/06/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

In machine learning, knowing the impact of a given datum on model training is a fundamental task referred to as Data Valuation. Building on previous works from the literature, we have designed a novel canonical decomposition allowing practitioners to analyze any data valuation method as the combination of two parts: a utility function that captures characteristics from a given model and an aggregation procedure that merges such information. We also propose to use Gaussian Processes as a means to easily access the utility function on ``sub-models'', which are models trained on a subset of the training set. The strength of our approach stems from both its theoretical grounding in Bayesian theory, and its practical reach, by enabling fast estimation of valuations thanks to efficient update formulae.

Citations

• Tanimoto Random Features for Scalable Molecular Machine Learning
• LAVA: Data Valuation without Pre-Specified Learning Algorithms
• Beta Shapley: a Unified and Noise-reduced Data Valuation Framework for Machine Learning
• Uncertainty Quantification and Experimental Design for Large-Scale Linear Inverse Problems under Gaussian Process Priors
• Retiring Adult: New Datasets for Fair Machine Learning
• Fast calculation of Gaussian Process multiple-fold cross-validation residuals and their covariances
• A Distributional Framework for Data Valuation
• Data Shapley: Equitable Valuation of Data for Machine Learning
• Exact Gaussian Processes on a Million Data Points
• Towards Efficient Data Valuation Based on the Shapley Value
• Gaussian Processes for Big Data
• Practical Bayesian Optimization of Machine Learning Algorithms
• Calculations of Sobol indices for the Gaussian process metamodel
• Principles of geostatistics

Related

• DUPRE: Data Utility Prediction for Efficient Data Valuation
• EcoVal: An Efficient Data Valuation Framework for Machine Learning
• Data Shapley: Equitable Valuation of Data for Machine Learning
• CHG Shapley: Efficient Data Valuation and Selection towards Trustworthy Machine Learning
• Data Appraisal Without Data Sharing