Research map / Architectures and efficiency / CNN design
Super Sparse Convolutional Neural 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 |
|---|---|---|---|
| replace Shift-Net with SSC kernel (SSC-Net-3-9, 0.56M vs 1.76M params) | mobile CNN on CIFAR10 · CIFAR10 | accuracy (%): 92.79 -> 93.93 | Helped |
| replace Shift-Net-18-6 with SSC-Net-6-9 at similar size | mobile CNN · CIFAR100 | accuracy (%): 72.56 -> 75.99 | Helped |
| replace ResNet-4-2 / Shift-Net-4-2 with SSC-Net-4-9 at 1.0M params | mobile CNN · ImageNet-32x32 | Top-1 accuracy (%): 41.47 -> 45.91 | Helped |
| replace Shift-Net-4-8 with SSC-Net-4-18 at 3.5M params | mobile CNN · ImageNet-32x32 | Top-1 accuracy (%): 51.85 -> 53.46 | Helped |
| use SSC-Net with 2x-2.5x fewer params instead of ResNet | ImageNet-32x32 · ImageNet-32x32 | Top-1 accuracy (%): 43.08 -> 43.88 | Helped |
| increase width/size of Shift-Net from 18-6 to 18-9 (1.18M to 1.76M) | Shift-Net · CIFAR10 | accuracy (%): 93.17 -> 92.79 | Hurt |
| increase SSC-Net size from 2-9 to 3-9 | SSC-Net · CIFAR10 | accuracy (%): 93.24 -> 93.93 | Helped |
| SSC-Net-4-18 vs IGCV3-D 1.0x and DenseNet-BC | mobile CNN · CIFAR100 | accuracy (%): 77.95 -> 77.67 | Hurt |
Closest papers by meaning
- Sparsity May Cry: Let Us Fail (Current) Sparse Neural Networks Together! (ICLR 2023 · similarity 0.85)
- All You Need Is a Few Shifts: Designing Efficient Convolutional Neural Networks for Image Classification (CVPR 2019 · similarity 0.84)
- When to Use Convolutional Neural Networks for Inverse Problems (CVPR 2020 · similarity 0.84)
- Exploring Sparsity in Image Super-Resolution for Efficient Inference (CVPR 2021 · similarity 0.84)
- SSN: Learning Sparse Switchable Normalization via SparsestMax (CVPR 2019 · similarity 0.84)
- Lightweight Image Super-Resolution via Flexible Meta Pruning (ICML 2024 · similarity 0.83)
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.