Research map / Architectures and efficiency / RNNs and normalization
Collegial Ensembles
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 |
|---|---|---|---|
| Replace the ResNeXt-29 3×64d baseline with the equal-parameter optimally smooth 37×10d ensemble | ResNeXt-29 · CIFAR-10 | test error (%): 3.96 -> 3.82 | Helped |
| Replace the ResNeXt-29 3×64d baseline with the equal-parameter optimally smooth 37×10d ensemble | ResNeXt-29 · CIFAR-100 | test error (%): 18.57 -> 17.94 | Helped |
| Use the compact 10×10d ensemble instead of the ResNeXt-29 1×128d baseline | ResNeXt-29 · CIFAR-10 | test error (%): 4.75 -> 4.70 | Helped |
| Use the compact 10×10d ensemble instead of the ResNeXt-29 1×128d baseline | ResNeXt-29 · CIFAR-100 | test error (%): 20.74 -> 20.84 | Hurt |
| Use the compact 10×10d ensemble instead of the ResNeXt-29 1×128d baseline | ResNeXt-29 · CIFAR-10 | test error (%): 4.08 -> 4.21 | Helped |
| Use the compact 10×10d ensemble instead of the ResNeXt-29 1×128d baseline | ResNeXt-29 · CIFAR-100 | test error (%): 19.39 -> 19.67 | Hurt |
| Use the optimally smooth 12×10d ensemble in place of ResNeXt-50 32×4d | ResNet-50 · ImageNet | Top-1 error: 22.42 -> 22.37 | Helped |
| Use the optimally smooth 12×10d ensemble in place of ResNeXt-50 32×4d | ResNet-50 · ImageNet | Top-5 error: 6.36 -> 6.30 | Helped |
| Use the optimally smooth 12×10d ensemble in place of ResNeXt-101 32×4d | ResNet-101 · ImageNet | Top-1 error: 21.01 -> 21.16 | Hurt |
| Use the optimally smooth 12×10d ensemble in place of ResNeXt-101 32×4d | ResNet-101 · ImageNet | Top-5 error: 5.72 -> 5.74 | Hurt |
| Replace the ResNeXt-29 3×64d baseline with the optimally smooth 37×10d ensemble | ResNeXt-29 · ImageNet 32×32 | Top-1 error: 39.58 -> 38.41 | Helped |
| Replace the ResNeXt-29 3×64d baseline with the optimally smooth 37×10d ensemble | ResNeXt-29 · ImageNet 32×32 | Top-5 error: 17.09 -> 16.13 | Helped |
Closest papers by meaning
- More or Less: When and How to Build Convolutional Neural Network Ensembles (ICLR 2021 · similarity 0.88)
- Theoretical Limitations of Ensembles in the Age of Overparameterization (ICML 2025 · similarity 0.88)
- Snapshot Ensembles: Train 1, Get M for Free (ICLR 2017 · similarity 0.88)
- Equivariant Neural Tangent Kernels (ICML 2025 · similarity 0.86)
- Ex Uno Pluria: Insights on Ensembling in Low Precision Number Systems (NeurIPS 2024 · similarity 0.85)
- Emergent Equivariance in Deep Ensembles (ICML 2024 · similarity 0.85)
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.