Research map / Architectures and efficiency / RNNs and normalization
Learning Deep ResNet Blocks Sequentially using Boosting Theory
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 |
|---|---|---|---|
| Train residual blocks sequentially with BoostResNet instead of end-to-end backpropagation | CNN-ResNet · SVHN | test accuracy: 83% -> 93.8% | Helped |
| Train residual blocks sequentially with BoostResNet instead of end-to-end backpropagation | CNN-ResNet · SVHN | training accuracy: 85% -> 96.9% | Helped |
| Use BoostResNet initialization followed by e2eBP refinement instead of e2eBP alone | CNN-ResNet · SVHN | test accuracy: 96.8% -> 96.8% | No clear effect |
| Train residual blocks sequentially with BoostResNet instead of end-to-end backpropagation | CNN-ResNet · CIFAR-10 | test accuracy: 80% -> 82.1% | Helped |
| Use BoostResNet initialization followed by e2eBP refinement instead of e2eBP alone | CNN-ResNet · CIFAR-10 | test accuracy: 90.0% -> 88.1% | Hurt |
| Train residual blocks sequentially with BoostResNet instead of end-to-end backpropagation | CNN-ResNet · CIFAR-10 | number of gradient updates: 3 × 10⁹ -> 3 × 10⁹ | Helped |
| Train residual blocks sequentially with BoostResNet instead of end-to-end backpropagation | MLP-ResNet · MNIST | training accuracy | Mixed |
Closest papers by meaning
- Functional Gradient Boosting based on Residual Network Perception (ICML 2018 · similarity 0.85)
- Towards Understanding the Importance of Shortcut Connections in Residual Networks (NeurIPS 2019 · similarity 0.85)
- Are deep ResNets provably better than linear predictors? (NeurIPS 2019 · similarity 0.85)
- Residual Connections Encourage Iterative Inference (ICLR 2018 · similarity 0.85)
- Generalization of Scaled Deep ResNets in the Mean-Field Regime (ICLR 2024 · similarity 0.83)
- A Rescaling-Invariant Lipschitz Bound Based on Path-Metrics for Modern ReLU Network Parameterizations (ICML 2025 · similarity 0.82)
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.