Research map / Learning theory and algorithms / Prediction and decision losses
Robust Data Valuation with Weighted Banzhaf Values
Weida Li, Yaoliang Yu
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 best weighted Banzhaf value instead of Banzhaf | semi-value · MNIST | ranking consistency: 0.581 ± 0.075 -> 0.680 ± 0.120 | Improved |
| Use best weighted Banzhaf value instead of Banzhaf | semi-value · gas | ranking consistency: 0.729 ± 0.019 -> 0.959 ± 0.009 | Improved |
| Use weighted Banzhaf instead of Shapley | semi-value · 2dplanes | ranking consistency: 0.256 ± 0.083 -> 0.950 ± 0.012 | Improved |
| Use weighted Banzhaf instead of Beta Shapley (Beta(16, 1)) | semi-value · FMNIST | ranking consistency: 0.017 ± 0.092 -> 0.670 ± 0.042 | Improved |
| Use weighted Banzhaf with parameter 0.00 on bioresponse | semi-value · bioresponse | ranking consistency: 0.948 ± 0.003 -> 0.997 ± 0.000 | Improved |
| Use best weighted Banzhaf instead of Banzhaf on letter | semi-value · letter | ranking consistency: 0.966 ± 0.004 | No clear effect |
| Use weighted Banzhaf instead of Banzhaf | semi-value · MNIST noisy label detection | F1: 0.625 ± 0.040 | No clear effect |
| Use weighted Banzhaf instead of Banzhaf | semi-value · 2dplanes noisy label detection | F1: 0.346 ± 0.014 -> 0.609 ± 0.059 | Improved |
| Use weighted Banzhaf instead of Shapley | semi-value · cpu noisy label detection | F1: 0.390 ± 0.027 -> 0.772 ± 0.017 | Improved |
| Use weighted Banzhaf instead of Shapley | semi-value · bank-marketing noisy label detection | F1: 0.368 ± 0.025 -> 0.528 ± 0.015 | Improved |
| Choose small Banzhaf parameter (0.05) for noisy label detection | weighted Banzhaf · most datasets | F1 | Mixed |
| Pick the semi-value best at ranking consistency for label detection | semi-value · all datasets | ranking vs F1 | Hurt |
Closest papers by meaning
- One Sample Fits All: Approximating All Probabilistic Values Simultaneously and Efficiently (NeurIPS 2024 · similarity 0.84)
- DeRDaVa: Deletion-Robust Data Valuation for Machine Learning (AAAI 2024 · similarity 0.83)
- Non-Asymptotic and Non-Lipschitzian Bounds on Optimal Values in Stochastic Optimization Under Heavy Tails (ICML 2025 · similarity 0.83)
- Distributionally Robust Data Valuation (ICML 2024 · similarity 0.83)
- KAIROS: Scalable Model-Agnostic Data Valuation (NeurIPS 2025 · similarity 0.82)
- Regularized least squares learning with heavy-tailed noise is minimax optimal (NeurIPS 2025 · similarity 0.82)
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.
