Shapley Values: Paired-Sampling Approximations
Michael Mayer, Mario V. Wüthrich · 2025
Read Shapley Values: Paired-Sampling Approximations by Michael Mayer, Mario V. Wüthrich (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-2508-12947.txt.
Description
Michael Mayer and Mario Wüthrich study Shapley values, a fair way to split credit among inputs. Their 2025 paper comes from game theory's tool kit, now widely used to explain machine learning forecasts. Each input feature gets credit for its share of a prediction. Exact credits are costly to compute, so practitioners sample instead.
Two sampling schemes dominate: one built on kernels and one on permutations. The authors prove both schemes' estimates settle into a bell-shaped spread as samples grow. They also test paired sampling, a tweak that draws samples in matched pairs. Paired sampling gives exact answers when features interact at most in pairs. The permutation version also recovers totals exactly, while the kernel version does not.
Catalog record
| CREATOR | Michael Mayer, Mario V. Wüthrich |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-12947