Research map / Image generation / Autoregressive and latent generation
The Intrinsic Dimension of Images and Its Impact on Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Increase intrinsic dimension while holding extrinsic dimension fixed | ResNet-18 on synthetic BigGAN image classification · Synthetic basenji-versus-beagle data | sample complexity | Hurt |
| Increase extrinsic dimension while holding intrinsic dimension fixed | ResNet-18 on synthetic BigGAN image classification · Synthetic basenji-versus-beagle data | sample complexity | No clear effect |
| Use datasets with higher estimated intrinsic dimension | ResNet-18 binary classification · MNIST, SVHN, CIFAR-10, and ImageNet | sample complexity | Mixed |
| Add uniformly sampled noise of increasing dimension | CIFAR-10 · Noisy CIFAR-10 binary classification | sample complexity | Hurt |
| Add uniformly sampled noise of increasing dimension | CIFAR-10 · Noisy CIFAR-10 | intrinsic dimension estimate: 27.6 -> 42.5 -> 78.3 -> 110.2 -> 153.4 -> 196.6 | Helped |
| Add more geometric transformations to the data | FONTS classification · FONTS-0 through FONTS-4 | sample complexity | Hurt |
| Add more geometric transformations to the data | FONTS · FONTS-0 through FONTS-4 | intrinsic dimension estimate: 3.2 (0.1) -> 5.9 (0.1) -> 7.8 (0.3) -> 8.8 (0.4) -> 9.4 (0.4) | Helped |
| Increase k in the MLE estimator (ablation) | MLE intrinsic dimension estimation · Synthetic basenji images | estimated intrinsic dimension | Mixed |
| Use GeoMLE, TwoNN, or kNN Graph Distance instead of MLE | Intrinsic dimension estimation on image datasets · basenji_10 and popular image datasets | estimation accuracy | Hurt |
Closest papers by meaning
- The Effect of Intrinsic Dataset Properties on Generalization: Unraveling Learning Differences Between Natural and Medical Images (ICLR 2024 · similarity 0.88)
- On the Intrinsic Dimensionality of Image Representations (CVPR 2019 · similarity 0.88)
- SMaRt: Improving GANs with Score Matching Regularity (ICML 2024 · similarity 0.85)
- A Geometric View of Data Complexity: Efficient Local Intrinsic Dimension Estimation with Diffusion Models (NeurIPS 2024 · similarity 0.85)
- A Wiener Process Perspective on Local Intrinsic Dimension Estimation Methods (AAAI 2025 · similarity 0.85)
- Dimension-Independent Rates for Structured Neural Density Estimation (ICML 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.