Research map / Learning theory and algorithms / Kernels and regression theory
Sample based Explanations via Generalized Representers
Che-Ping Tsai, Chih-Kuan Yeh, Pradeep Kumar Ravikumar
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 tracking global importance instead of surrogate derivative (ablation) | NTK-final generalized representer · MNIST | AUC-DEL-: 1.88 -> 3.52 | Improved |
| use tracking global importance instead of surrogate derivative (ablation) | NTK-final generalized representer · CIFAR-10 | AUC-DEL-: 2.27 -> 3.26 | Improved |
| use tracking global importance instead of target derivative (ablation) | NTK-final generalized representer · Toxicity | AUC-DEL-: 1.10 -> 2.08 | Improved |
| use tracking global importance instead of target derivative (ablation) | NTK-final generalized representer · AGnews | AUC-DEL-: 1.88 -> 2.56 | Improved |
| use target derivative instead of surrogate derivative (ablation) | NTK-final generalized representer · MNIST | AUC-DEL-: 1.88 -> 2.41 | Improved |
| use target derivative instead of surrogate derivative (ablation) | NTK-final generalized representer · CIFAR-10 | AUC-DEL-: 2.27 -> 2.81 | Improved |
| use NTK-final kernel instead of NTK-init kernel (ablation) | tracking representer · CIFAR-10 | AUC-DEL-: 1.35 -> 3.26 | Improved |
| use NTK-final kernel instead of NTK-init kernel (ablation) | tracking representer · MNIST | AUC-DEL-: 3.18 -> 3.52 | Mixed |
| use NTK-final kernel instead of last-layer kernel (ablation) | tracking representer · CIFAR-10 | AUC-DEL-: 2.26 -> 3.26 | Improved |
| use NTK-final kernel instead of NTK-middle kernel (ablation) | tracking representer · AGnews | AUC-DEL-: 2.54 -> 2.56 | No clear effect |
| use influence-function kernel instead of NTK-final kernel (ablation) | tracking representer · CIFAR-10 | AUC-DEL-: 3.26 -> 3.46 | Improved |
| use influence-function kernel (last layer only) instead of NTK-final kernel (ablation) | tracking representer · Toxicity | AUC-DEL-: 2.08 -> 0.42 | Hurt |
Closest papers by meaning
- Representer Point Selection for Explaining Regularized High-dimensional Models (ICML 2023 · similarity 0.87)
- Data-centric Prediction Explanation via Kernelized Stein Discrepancy (ICLR 2025 · similarity 0.86)
- Efficient and Accurate Explanation Estimation with Distribution Compression (ICLR 2025 · similarity 0.85)
- Additive Models Explained: A Computational Complexity Approach (NeurIPS 2025 · similarity 0.85)
- GEFA: A General Feature Attribution Framework Using Proxy Gradient Estimation (ICML 2025 · similarity 0.85)
- Explainable Neural Networks with Guarantee: A Sparse Estimation Approach (AAAI 2025 · similarity 0.84)
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.
