On the Usage of Gaussian Process for Efficient Data Valuation
Clément Bénesse, Patrick Mesana, Athénaïs Gautier, Sébastien Gambs · 2025
Read On the Usage of Gaussian Process for Efficient Data Valuation by Clément Bénesse, Patrick Mesana, Athénaïs Gautier, Sébastien Gambs (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-04026.txt.
Description
Clément Bénesse and co-authors ask how much each piece of training data is worth to a learning model. Their 2025 paper calls this task data valuation. Past work is hard to use fast, so the team looks for a way to judge data points without retraining large models again and again.
They split any valuation method into two parts: a utility function tied to the model, and a step that merges the scores. Then they use Gaussian processes to read the utility from small models trained on parts of the data. Fast update rules make the scores cheap to get. The authors ground the plan in Bayes based theory and aim it at daily practice.
Catalog record
| CREATOR | Clément Bénesse, Patrick Mesana, Athénaïs Gautier, Sébastien Gambs |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 87 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-04026