Research map / Learning theory and algorithms / Prediction and decision losses
$\texttt{ShaplEIG}$: Bayesian Experimental Design for Shapley Value Estimation
David Rundel, Fabian Fumagalli, Maximilian Muschalik, Bernd Bischl, Matthias Feurer
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| select coalitions by EIG instead of fixed-distribution sampling (Kernel SHAP, Leverage SHAP, permutation, Regression MSR) | ShaplEIG · costly games (FI, DV, HPI, LE), low budget | SV MSE | Improved |
| replace EIG selection with random coalition selection (GP + Random) (ablation) | GP surrogate with Hamming kernel · all tasks | SV MSE | Improved |
| add EIG selection instead of uncertainty sampling (GP + US) (ablation) | GP surrogate with Hamming kernel · all tasks | SV MSE | Improved |
| add EIG selection instead of leverage score sampling (GP + Leverage) (ablation) | GP surrogate with Hamming kernel · all tasks | SV MSE | Improved |
| EIG selection vs GP ablations on large local explanation games (ablation) | ShaplEIG · large LE games (p up to 101) | SV MSE | Mixed |
| EIG selection on a symmetric game (|S|^2) | ShaplEIG · 3-player symmetric toy game | coalition selection | No clear effect |
| EIG selection on an asymmetric game | ShaplEIG · 3-player asymmetric toy game | SV MSE: 0.024 -> 5.485e-06 (vs 0.024 -> 8.75e-4 for size-based sampling) | Improved |
| refit GP hyperparameters only on a schedule for larger games (p > 16) | ShaplEIG · large games | SV MSE | No clear effect |
Closest papers by meaning
- Regression-adjusted Monte Carlo Estimators for Shapley Values and Probabilistic Values (NeurIPS 2025 · similarity 0.89)
- Provably Accurate Shapley Value Estimation via Leverage Score Sampling (ICLR 2025 · similarity 0.88)
- DU-Shapley: A Shapley Value Proxy for Efficient Dataset Valuation (NeurIPS 2024 · similarity 0.88)
- shapiq: Shapley Interactions for Machine Learning (NeurIPS 2024 · similarity 0.88)
- A Unified Framework for Provably Efficient Algorithms to Estimate Shapley Values (NeurIPS 2025 · similarity 0.87)
- Rethinking Data Shapley for Data Selection Tasks: Misleads and Merits (ICML 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
