Research map / Architectures and efficiency / RNNs and normalization
Meta-Learning Sparse Implicit Neural Representations
Jaeho Lee, Jihoon Tack, Namhoon Lee, Jinwoo Shin
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 Meta-SparseINR instead of Dense-Narrow at a similar parameter count | SIREN · CelebA | PSNR: \approx 20.52 -> \approx 26.73 | Improved |
| Use Meta-SparseINR instead of random pruning | FFN · CelebA, 41.0% surviving parameters | PSNR: 21.54\stdv{0.17} -> 23.91\stdv{0.23} | Improved |
| Use Meta-SparseINR instead of Dense-Narrow | FFN · CelebA, approximately 41% surviving parameters | PSNR: 19.39\stdv{2.37} -> 23.91\stdv{0.23} | Improved |
| Use Meta-SparseINR instead of random pruning | FFN · CelebA, 10.74% surviving parameters | PSNR: 15.88\stdv{0.23} -> 18.41\stdv{0.38} | Improved |
| Use Meta-SparseINR instead of Dense-Narrow | FFN · CelebA, approximately 11–13% surviving parameters | PSNR: 14.95\stdv{1.76} -> 18.41\stdv{0.38} | Improved |
| Replace magnitude pruning with random pruning (ablation) | Meta-SparseINR · CelebA, FFN with 80.0% surviving parameters | PSNR: 24.88\stdv{1.62} vs. 25.15\stdv{0.20} | No clear effect |
| Replace meta-learned weights with standard random initialization (ablation) | Dense-Narrow · CelebA, FFN with 80.8% surviving-parameter counterpart | PSNR: 22.27\stdv{2.55} -> 15.82\stdv{0.04} | Hurt |
| Use winning-ticket pruning at initialization instead of trained initial weights (ablation) | Sparse INR · 512×512 natural images | PSNR | No clear effect |
| Use Reptile instead of MAML for meta-learning (ablation) | Meta-SparseINR · CelebA | PSNR: ≈ 2 | Hurt |
| Use Meta-SparseINR instead of JPEG compression | Image compression · Image compression comparison | bits per pixel | Hurt |
Closest papers by meaning
- Transformers As Meta-Learners for Implicit Neural Representations (ECCV 2022 · similarity 0.90)
- SINR: Sparsity Driven Compressed Implicit Neural Representations (CVPR 2025 · similarity 0.88)
- From Flat to Hierarchical: Extracting Sparse Representations with Matching Pursuit (NeurIPS 2025 · similarity 0.86)
- MetaNeRV: Meta Neural Representations for Videos with Spatial-Temporal Guidance (AAAI 2025 · similarity 0.86)
- MIRE: Matched Implicit Neural Representations (CVPR 2025 · similarity 0.86)
- Beyond Matryoshka: Revisiting Sparse Coding for Adaptive Representation (ICML 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.
