Research map / Architectures and efficiency / CNN design
Efficient Sparse-Winograd 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 |
|---|---|---|---|
| Move ReLU to the Winograd domain and prune transformed weights | VGG-nagadomi · CIFAR-10 | overall workload reduction: 4.7× -> 10.4×; 3.5× -> 10.4× | Helped |
| Move ReLU to the Winograd domain and prune transformed weights | ConvPool-CNN-C · CIFAR-100 | overall workload reduction: 3.4× -> 6.8×; 3.1× -> 6.8× | Helped |
| Move ReLU to the Winograd domain and prune transformed weights | ResNet-18 variation · ImageNet 2012 | overall workload reduction: 4.7× -> 10.8×; 4.2× -> 10.8× | Helped |
| Prune weights in the Winograd domain without moving ReLU there (ablation) | Conventional Winograd CNN · CIFAR-10 | overall workload reduction: 4.7× -> 3.5× | Hurt |
| Prune weights in the Winograd domain without moving ReLU there (ablation) | Conventional Winograd CNN · CIFAR-100 | overall workload reduction: 3.4× -> 3.1× | Hurt |
| Prune weights in the Winograd domain without moving ReLU there (ablation) | Conventional Winograd CNN · ImageNet 2012 | overall workload reduction: 4.7× -> 4.2× | Hurt |
| Move ReLU to the Winograd domain and prune weights more aggressively (ablation) | VGG-nagadomi · CIFAR-10 | weight density at similar accuracy: 60% -> 40% | Helped |
| Move ReLU to the Winograd domain and prune weights more aggressively (ablation) | ConvPool-CNN-C · CIFAR-100 | weight density at similar accuracy: 60% -> 60%; 70% -> 60% | Mixed |
| Move ReLU to the Winograd domain and prune weights more aggressively (ablation) | ResNet-18 variation · ImageNet 2012 | weight density at similar accuracy: 60% -> 30%/35%; 50% -> 30%/35% | Helped |
Closest papers by meaning
- DWM: A Decomposable Winograd Method for Convolution Acceleration (AAAI 2020 · similarity 0.90)
- Towards Efficient and Accurate Winograd Convolution via Full Quantization (NeurIPS 2023 · similarity 0.88)
- Faster CNNs with Direct Sparse Convolutions and Guided Pruning (ICLR 2017 · similarity 0.87)
- Accelerating Sparse Convolution with Column Vector-Wise Sparsity (NeurIPS 2022 · similarity 0.87)
- Channel Balancing for Accurate Quantization of Winograd Convolutions (CVPR 2022 · similarity 0.87)
- Data-Free Group-Wise Fully Quantized Winograd Convolution via Learnable Scales (CVPR 2025 · similarity 0.86)
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.