Research map / Architectures and efficiency / RNNs and normalization
PolyNet: A Pursuit of Structural Diversity in Very Deep Networks
Check 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 Inception residual units with PolyInception modules (ablation) | IR 3-6-3 · ILSVRC 2012 validation set, single-crop | Top-5 accuracy | Helped |
| Use a mixed sequence of 3-way, mpoly-3, and poly-3 modules instead of uniform modules (ablation) | IR 6-12-6, stage B · ILSVRC 2012 validation set, single-crop | Top-5 error | Helped |
| Use poly-3 instead of mpoly-3 (ablation) | IR 3-6-3, stage B · ILSVRC 2012 validation set, single-crop | Top-5 accuracy | No clear effect |
| Increase ResNet depth from ResNet-269 to ResNet-500 (ablation) | ResNet · ILSVRC 2012 validation set, single-crop | Top-5 error: 4.9% -> 4.8% | Helped |
| Increase ResNet depth from ResNet-269 to ResNet-500 (ablation) | ResNet · ILSVRC 2012 validation set, multi-crop | Top-5 error: 3.55 -> 3.63 | Hurt |
| Increase network depth rather than structural diversity at larger compute budgets (ablation) | Inception-ResNet-v2 and PolyNet · ILSVRC 2012 validation set, single-crop | Top-5 accuracy | Mixed |
Closest papers by meaning
- Revisiting ResNets: Improved Training and Scaling Strategies (NeurIPS 2021 · similarity 0.87)
- Regularization of Polynomial Networks for Image Recognition (CVPR 2023 · similarity 0.87)
- Rethinking the Inception Architecture for Computer Vision (CVPR 2016 · similarity 0.86)
- Unveiling the Unseen: Identifiable Clusters in Trained Depthwise Convolutional Kernels (ICLR 2024 · similarity 0.86)
- Depthwise Hyperparameter Transfer in Residual Networks: Dynamics and Scaling Limit (ICLR 2024 · similarity 0.85)
- Principled Architecture-aware Scaling of Hyperparameters (ICLR 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.