Research map / Learning theory and algorithms / Prediction and decision losses
SHAP-IQ: Unified Approximation of any-order Shapley Interactions
Fabian Fumagalli, Maximilian Muschalik, Patrick Kolpaczki, Eyke Hüllermeier, Barbara Hammer
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 |
|---|---|---|---|
| use SHAP-IQ instead of permutation-based baseline for SII | SII approximation · IMDB DistilBERT LM, orders 1-3 | MSE | Improved |
| use SHAP-IQ instead of baseline for SII | SII approximation · LM and ResNet18 image model | Prec@10 | Improved |
| use SHAP-IQ without sampling (only c_k0 term) for n-SII visualization (ablation) | SHAP-IQ · n-SII visualization, s0=4, LM | accuracy of n-SII estimate vs ground truth | Improved |
| use SHAP-IQ instead of PB baseline for STI | STI top-order approximation · LM, s0=3 | MSE | Improved |
| use SHAP-IQ instead of KB approximation for FSI | FSI approximation · LM, s0=3 | MSE | Hurt |
| use SHAP-IQ instead of baselines (PB for SII/STI, KB for FSI) | SII, STI, FSI pairwise interactions · SOUM d=30, s0=2 | MSE and Prec@K | Improved |
| use KB approximation instead of PB approximation | baseline interaction approximators · LM | approximation quality | Mixed |
| increase number of model evaluations K | SHAP-IQ and baselines · LM, 16 words | runtime | Hurt |
| use SHAP-IQ instead of baseline regarding runtime overhead | SHAP-IQ · LM, 16 words, five runs | runtime | No clear effect |
Closest papers by meaning
- KernelSHAP-IQ: Weighted Least Square Optimization for Shapley Interactions (ICML 2024 · similarity 0.92)
- shapiq: Shapley Interactions for Machine Learning (NeurIPS 2024 · similarity 0.92)
- Beyond TreeSHAP: Efficient Computation of Any-Order Shapley Interactions for Tree Ensembles (AAAI 2024 · similarity 0.91)
- Exact Computation of Any-Order Shapley Interactions for Graph Neural Networks (ICLR 2025 · similarity 0.88)
- SHAP values via sparse Fourier representation (NeurIPS 2025 · similarity 0.88)
- Towards Trustable SHAP Scores (AAAI 2025 · similarity 0.88)
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.
