Research map / Architectures and efficiency / CNN design
S3Pool: Pooling With Stochastic Spatial Sampling
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 deterministic spatial downsampling with S3Pool-16-8 | NIN · CIFAR-10 | test error: 9.34 -> 7.71 | Helped |
| Replace deterministic spatial downsampling with S3Pool-16-8 | ResNet · CIFAR-10 | test error: 7.72 -> 7.09 | Helped |
| Replace deterministic spatial downsampling with S3Pool-16-8 | NIN · CIFAR-100 | test error: 32.36 -> 30.90 | Helped |
| Replace deterministic spatial downsampling with S3Pool-16-8 | ResNet · CIFAR-100 | test error: 30.88 -> 29.36 | Helped |
| Use S3Pool-96-48-24-12 instead of the ResNet baseline | 18-layer ResNet · STL-10 | test error: 39.84 -> 24.06 | Helped |
| Increase grid sizes from 16-8 to 32-16 (ablation) | S3Pool with ResNet · CIFAR-10 | test error: 8.86 -> 9.32 | Hurt |
| Reduce grid sizes from 16-8 to 8-8 (ablation) | S3Pool with ResNet · CIFAR-10 | test error: 8.86 -> 10.14 | Hurt |
| Reduce grid sizes from 16-8 to 8-4 (ablation) | S3Pool with ResNet · CIFAR-10 | test error: 8.86 -> 11.04 | Hurt |
| Reduce grid sizes from 16-8 to 4-4 (ablation) | S3Pool with ResNet · CIFAR-10 | test error: 8.86 -> 11.02 | Hurt |
| Reduce grid sizes from 16-8 to 2-2 (ablation) | S3Pool with ResNet · CIFAR-10 | test error: 8.86 -> 13.01 | Hurt |
| Use grid sizes 48, 24, 12, 6 instead of 96, 48, 24, 12 (ablation) | S3Pool with 18-layer ResNet · STL-10 | test error: 24.06 -> 25.36 | Hurt |
| Use grid sizes 24, 12, 6, 4 instead of 96, 48, 24, 12 (ablation) | S3Pool with 18-layer ResNet · STL-10 | test error: 24.06 -> 29.21 | Hurt |
Closest papers by meaning
- Detail-Preserving Pooling in Deep Networks (CVPR 2018 · similarity 0.90)
- Gaussian-Based Pooling for Convolutional Neural Networks (NeurIPS 2019 · similarity 0.90)
- Wavelet Pooling for Convolutional Neural Networks (ICLR 2018 · similarity 0.90)
- Learning with convolution and pooling operations in kernel methods (NeurIPS 2022 · similarity 0.88)
- Beyond Skip Connection: Pooling and Unpooling Design for Elimination Singularities (AAAI 2025 · similarity 0.86)
- SkipPool: Improved Sparse Hierarchical Graph Pooling with Differentiable Exploration (AAAI 2025 · 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.