Research map / Adversarial robustness and security / Certified and empirical robustness
Interpolation between Residual and Non-Residual Networks
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 |
|---|---|---|---|
| Add interpolated residual connections | ResNet-110 · CIFAR-10-C stochastic noise, Avg. | accuracy: 53.70 -> 72.67 | Helped |
| Add interpolated residual connections | ResNet-110 · CIFAR-10 test | accuracy: 93.58 -> 92.28 | Hurt |
| Add interpolated residual connections | ResNet-110 · CIFAR-10, IFGSM 2/255 | accuracy: 5.93 -> 32.05 | Helped |
| Add interpolated residual connections | ResNet-110 · CIFAR-100, PGD 4/255 | accuracy: 2.11 -> 4.81 | Helped |
| Replace ReLU with sigmoid for interpolation coefficients (ablation) | In-ResNet-110 · CIFAR-10-C stochastic noise, Avg. | accuracy: 72.67 -> 55.04 | Hurt |
| Make interpolation coefficients input-dependent with gating (ablation) | In-ResNet-110 · CIFAR-10-C stochastic noise, Avg. | accuracy: 72.67 -> 54.53 | Hurt |
| Use input-dependent gating and sigmoid coefficients (ablation) | In-ResNet-110 · CIFAR-10, IFGSM 2/255 | accuracy: 32.05 -> 21.89 | Hurt |
| Increase interpolation-coefficient initialization from U[0.20, 0.25] to U[0.30, 0.40] (ablation) | In-ResNet-110 · CIFAR-10-C stochastic noise, Avg. | accuracy: 72.67 -> 79.35 | Helped |
| Increase interpolation-coefficient initialization from U[0.20, 0.25] to U[0.30, 0.40] (ablation) | In-ResNet-110 · CIFAR-10 test | accuracy: 92.28 -> 90.62 | Hurt |
| Ensemble predictions from 5 runs | In-ResNet-110 · CIFAR-10-C stochastic noise, Avg. | accuracy: 72.67 -> 75.86 | Helped |
Closest papers by meaning
- Scaling Properties of Deep Residual Networks (ICML 2021 · similarity 0.87)
- Multi-level Residual Networks from Dynamical Systems View (ICLR 2018 · similarity 0.86)
- Reversible Architectures for Arbitrarily Deep Residual Neural Networks (AAAI 2018 · similarity 0.86)
- Implicit regularization of deep residual networks towards neural ODEs (ICLR 2024 · similarity 0.86)
- Do Residual Neural Networks discretize Neural Ordinary Differential Equations? (NeurIPS 2022 · similarity 0.86)
- Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit (ICLR 2024 · 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.