Research map / Learning theory and algorithms / Sample complexity
Deep Network Approximation in Terms of Intrinsic Parameters
Zuowei Shen, Haizhao Yang, Shijun Zhang
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 |
|---|---|---|---|
| Train only Blocks 2 and 4 after pre-training all parameters on labels 0–4 | Proposed CNN · MNIST test | test accuracy: 0.9964 -> 0.9962 | Hurt |
| Train only Blocks 2 and 4 after pre-training all parameters on labels 0–4 | Proposed CNN · KMNIST test | test accuracy: 0.9832 -> 0.9744 | Hurt |
| Train only Blocks 2 and 4 after pre-training all parameters on labels 0–4 | Proposed CNN · FMNIST test | test accuracy: 0.9458 -> 0.9360 | Hurt |
| Train only Blocks 2 and 4 rather than all parameters | Proposed CNN · MNIST, KMNIST, FMNIST | number of training parameters: $2.2\times 10^6$ -> $8.9\times 10^4$ | Improved |
Closest papers by meaning
- How Much Over-parameterization Is Sufficient to Learn Deep ReLU Networks? (ICLR 2021 · similarity 0.89)
- ReLU Network with Width $d+\mathcal{O}(1)$ Can Achieve Optimal Approximation Rate (ICML 2024 · similarity 0.89)
- Deep Network Approximation: Beyond ReLU to Diverse Activation Functions (JMLR 2024 · similarity 0.89)
- Characterizing ResNet's Universal Approximation Capability (ICML 2024 · similarity 0.88)
- Compelling ReLU Networks to Exhibit Exponentially Many Linear Regions at Initialization and During Training (ICML 2025 · similarity 0.88)
- Minimum width for universal approximation using ReLU networks on compact domain (ICLR 2024 · similarity 0.88)
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.
