Research map / 3D vision / Face and shape
NTK-Guided Implicit Neural Teaching
Chen Zhang, Wei Zuo, Bingyang Cheng, Yikun Wang, Wei-Bin Kou, Yik-Chung Wu, Ngai Wong
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 NTK-guided sampling instead of full-coordinate training | SIREN 5x256 · Image fitting, time to PSNR 35 dB | time (s): 184.78 -> 102.88 | Improved |
| use NTK-guided sampling instead of full-coordinate training | SIREN 5x256 · Image fitting, 5000 iters | PSNR: 39.763 -> 39.085 | Hurt |
| use NTK-guided sampling instead of error-only INT | SIREN 5x256 · Image fitting, 250 iters | PSNR: 27.568 -> 28.956 | Improved |
| use NINT versus Expansive Supervision | SIREN 5x256 · Image fitting, 1000 iters | SSIM: 0.857 -> 0.841 | Hurt |
| add NINT to SIREN (ablation) | SIREN 5x256 · Image fitting, 2500 iters | PSNR: 33.693 -> 35.100 | Improved |
| add NINT to small SIREN 3x64 (ablation) | SIREN 3x64 · Image fitting, 500 iters | PSNR: 22.291 -> 22.023 | Hurt |
| add NINT to larger network (ablation) | SIREN 5x256 · Image fitting, time to PSNR 25 dB | time (s): 35.42 -> 22.16 | Improved |
| add NINT to various INR architectures (ablation) | FFN, FINER, GAUSS, PEMLP, SIREN, WIRE, MLP · Image fitting, time to PSNR 25 dB | time (s): WIRE 83.30 -> 47.23; FINER 6.90 -> 6.76 | Improved |
| add NINT to GAUSS at 20s (ablation) | GAUSS · Image fitting, 20s training | PSNR: 21.51 -> 21.46 | No clear effect |
| add NINT to FINER (ablation) | FINER · Image fitting, 120s training | PSNR: 33.01 -> 35.61 | Improved |
| decrease NTK decay rate lambda from 1.0 to 0.1 (ablation) | NINT · Image fitting, time to PSNR 30 dB | time (s): 25.09 -> 28.48 | Hurt |
| increase lambda from 1.0 to 2.0 (ablation) | NINT · Image fitting, time to PSNR 30 dB | time (s): 25.09 -> 24.82 | No clear effect |
Closest papers by meaning
- Nonparametric Teaching of Implicit Neural Representations (ICML 2024 · similarity 0.88)
- VI^3NR: Variance Informed Initialization for Implicit Neural Representations (CVPR 2025 · similarity 0.87)
- Neural Experts: Mixture of Experts for Implicit Neural Representations (NeurIPS 2024 · similarity 0.86)
- The Surprising Effectiveness of Infinite-Width NTKs for Characterizing and Improving Model Training (AAAI 2025 · similarity 0.86)
- Learning Transferable Features for Implicit Neural Representations (NeurIPS 2024 · similarity 0.85)
- MIRE: Matched Implicit Neural Representations (CVPR 2025 · similarity 0.85)
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.
