Research map / Label-efficient and robust learning / Domain adaptation and generalization
Why We Need New Benchmarks for Local Intrinsic Dimension Estimation
Piotr Tempczyk, Dominik Filipiak, Łukasz Garncarek, Ksawery Smoczyński, Adam Kurpisz
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 |
|---|---|---|---|
| map the manifold into image domain (IDR) instead of the original space (ablation) | LIDL · Gaussians | LID estimation MAE: 0.90 -> 12.24 | Hurt |
| map the manifold into image domain (IDR) instead of the original space (ablation) | LIDL · Uniform | LID estimation MAE: 0.88 -> 14.70 | Hurt |
| map the manifold into image domain (IDR) instead of the original space (ablation) | LIDL · Moon | LID estimation MAE: 0.24 -> 9.31 | Hurt |
| use convolutional networks (Glow, U-net) instead of feedforward (MAF, MLP) (ablation) | LIDL and FLIPD · IDR datasets | LID estimation MAE | Hurt |
| use ESS instead of neural estimators on non-uniform Gaussians | ESS · Gaussians (IDR) | LID estimation MAE: 0.07 vs 3.08, 12.24, 0.81 | Improved |
| use ESS instead of neural estimators on curved spheres | ESS · Spheres (IDR) | LID estimation MAE: 0.09 vs 3.24, 12.90, 1.83 | Improved |
| use NB instead of ESS on high-dimensional arrows manifolds | NB · Arrows (MS) | LID estimation MAE: 6.23 -> 440.82 | Hurt |
| apply stretching x^4 transformation to FMNIST (ablation) | FLIPD · FMNIST (stretched x4) | mean LID: 55.92 -> 86,355.13 | Hurt |
| apply stretching x^4 transformation to FMNIST (ablation) | ESS · FMNIST (stretched x4) | mean LID: 15.32 -> 17.98 | No clear effect |
| add 0 extra reflection dimensions to FMNIST (ablation) | FLIPD · FMNIST (add dim +0d) | mean LID: 55.92 -> 210.03 | Hurt |
| add 0 extra reflection dimensions to FMNIST (ablation) | ESS · FMNIST (add dim +0d) | mean LID: 15.32 -> 14.87 | No clear effect |
| use IDR with padding instead of original manifold (ablation) | LIDL · Gaussians | LID estimation MAE: 1.17 -> 22.14 | Hurt |
Closest papers by meaning
- A Wiener Process Perspective on Local Intrinsic Dimension Estimation Methods (AAAI 2025 · similarity 0.92)
- A Geometric View of Data Complexity: Efficient Local Intrinsic Dimension Estimation with Diffusion Models (NeurIPS 2024 · similarity 0.90)
- Intrinsic Dimension Correlation: uncovering nonlinear connections in multimodal representations (ICLR 2025 · similarity 0.85)
- Diffusion Models Encode the Intrinsic Dimension of Data Manifolds (ICML 2024 · similarity 0.84)
- Discriminating image representations with principal distortions (ICLR 2025 · similarity 0.83)
- Exploring Intrinsic Dimension for Vision-Language Model Pruning (ICML 2024 · similarity 0.83)
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.
