Research map / Architectures and efficiency / RNNs and normalization
Multi-level Residual Networks from Dynamical Systems View
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 |
|---|---|---|---|
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | ResNet-122 · CIFAR-10 | test error: 6.47% -> 6.56% | Hurt |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | ResNet-122 · CIFAR-10 | training time: 266m -> 154m | Helped |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | ResNet-122 · CIFAR-100 | test error: 26.74% -> 26.81% | Hurt |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | ResNet-122 · CIFAR-100 | training time: 266m -> 154m | Helped |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | ResNet-122 · STL-10 | test error: 25.16% -> 24.36% | Helped |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | ResNet-122 · STL-10 | training time: 266m -> 162m | Helped |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | WResNet-122 · CIFAR-10 | test error: 5.38% -> 5.46% | Hurt |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | WResNet-122 · CIFAR-10 | training time: 406m -> 239m | Helped |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | WResNet-122 · CIFAR-100 | test error: 23.11% -> 23.04% | Helped |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | WResNet-122 · CIFAR-100 | training time: 406m -> 237m | Helped |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | WResNet-122 · STL-10 | test error: 22.00% -> 22.65% | Hurt |
| Use multi-level training, progressing from 5-5-5 to 20-20-20 blocks | WResNet-122 · STL-10 | training time: 516m -> 307m | Helped |
Closest papers by meaning
- Residual Connections Encourage Iterative Inference (ICLR 2018 · similarity 0.90)
- Residual Alignment: Uncovering the Mechanisms of Residual Networks (NeurIPS 2023 · similarity 0.90)
- Revisiting ResNets: Improved Training and Scaling Strategies (NeurIPS 2021 · similarity 0.90)
- Dr2Net: Dynamic Reversible Dual-Residual Networks for Memory-Efficient Finetuning (CVPR 2024 · similarity 0.88)
- Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit (ICLR 2024 · similarity 0.88)
- Neural Collapse is Globally Optimal in Deep Regularized ResNets and Transformers (NeurIPS 2025 · similarity 0.87)
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.