Research map / Federated learning and privacy / Differential privacy
A Privacy-Friendly Approach to Data Valuation
Jiachen T. Wang, Yuqing Zhu, Yu-Xiang Wang, Ruoxi Jia, Prateek Mittal
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 DP-TKNN-Shapley instead of DP-KNN-Shapley without subsampling at eps=0.1 | DP data valuation · Mislabeled detection, 2dPlanes | AUROC: 0.49 -> 0.883 | Improved |
| Use DP-TKNN-Shapley instead of subsampled DP-KNN-Shapley at eps=0.1 | DP data valuation · Mislabeled detection, 2dPlanes | AUROC: 0.733 -> 0.883 | Improved |
| Use DP-TKNN-Shapley instead of subsampled DP-KNN-Shapley at eps=0.1 | DP data valuation · Noisy detection, CPU | AUROC: 0.709 -> 0.805 | Improved |
| Use DP-TKNN-Shapley instead of subsampled DP-KNN-Shapley at eps=0.1 | DP data valuation · Mislabeled detection, Phoneme | AUROC: 0.692 -> 0.816 | Improved |
| Add subsampling (q=0.01) to DP-KNN-Shapley | DP-KNN-Shapley · Mislabeled detection, 2dPlanes, eps=0.1 | AUROC: 0.49 -> 0.733 | Improved |
| Increase privacy budget eps from 0.1 to 1 | DP-TKNN-Shapley · Mislabeled detection, 2dPlanes | AUROC: 0.883 -> 0.913 | Improved |
| Increase privacy budget eps from 0.1 to 1 | DP-TKNN-Shapley · Mislabeled detection, Phoneme | AUROC: 0.816 -> 0.826 | Improved |
| Replace KNN-Shapley with TKNN-Shapley (non-private) | TKNN-Shapley · Mislabeled detection, CIFAR10 | AUROC: 0.991 -> 0.957 | Hurt |
| Replace KNN-Shapley with TKNN-Shapley (non-private) | TKNN-Shapley · Mislabeled detection, Pol | AUROC: 0.928 -> 0.871 | Hurt |
| Replace KNN-Shapley with TKNN-Shapley (non-private) | TKNN-Shapley · Mislabeled detection, Phoneme | AUROC: 0.873 -> 0.826 | Hurt |
| Replace KNN-Shapley with TKNN-Shapley (non-private) | TKNN-Shapley · Mislabeled detection, CPU | AUROC: 0.932 -> 0.946 | Improved |
| Replace KNN-Shapley with TKNN-Shapley (non-private) | TKNN-Shapley · 13 mislabeled detection datasets | AUROC | No clear effect |
Closest papers by meaning
- Data Shapley: Equitable Valuation of Data for Machine Learning (ICML 2019 · similarity 0.88)
- DU-Shapley: A Shapley Value Proxy for Efficient Dataset Valuation (NeurIPS 2024 · similarity 0.87)
- Certification for Differentially Private Prediction in Gradient-Based Training (ICML 2025 · similarity 0.86)
- Provable Privacy with Non-Private Pre-Processing (ICML 2024 · similarity 0.86)
- It's My Data Too: Private ML for Datasets with Multi-User Training Examples (ICML 2025 · similarity 0.86)
- Learning from End User Data with Shuffled Differential Privacy over Kernel Densities (ICLR 2025 · similarity 0.86)
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.
