Research map / Architectures and efficiency / CNN design
Network Sketching: Exploiting Binary Structure in Deep CNNs
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 |
|---|---|---|---|
| Use refined rather than direct approximation (ablation) | AlexNet sketch · ImageNet ILSVRC-2012 validation | top-1 accuracy (%): 54.4 -> 55.2 | Helped |
| Use refined rather than direct approximation (ablation) | AlexNet sketch · ImageNet ILSVRC-2012 validation | top-5 accuracy (%): 78.1 -> 78.8 | Helped |
| Use refined rather than direct approximation (ablation) | ResNet-18 type B sketch · ImageNet ILSVRC-2012 validation | top-1 accuracy (%): 67.3 -> 67.8 | Helped |
| Use refined rather than direct approximation (ablation) | ResNet-18 type B sketch · ImageNet ILSVRC-2012 validation | top-5 accuracy (%): 88.2 -> 88.4 | Helped |
| Replace direct binary convolutions with associative convolutions using an MST dependency tree (ablation) | AlexNet 3-bit refined sketch · Conv2_sketch | FADDs: $ sim$672M -> $ sim$265M | Helped |
| Replace direct binary convolutions with associative convolutions using an MST dependency tree (ablation) | AlexNet 3-bit refined sketch · Fc6_sketch | FADDs: $ sim$113M -> $ sim$49M | Helped |
| Use an MST rather than a stochastic dependency tree for associative convolutions (ablation) | AlexNet 3-bit refined sketch · Conv2_sketch | FADDs: $ sim$328M -> $ sim$265M | Helped |
| Use an MST rather than a stochastic dependency tree for associative convolutions (ablation) | AlexNet 3-bit refined sketch · Fc6_sketch | FADDs: $ sim$56M -> $ sim$49M | Helped |
| Use direct approximation rather than random binary basis (ablation) | AlexNet filter approximation · ImageNet ILSVRC-2012 learnable layers | accumulative energy | Helped |
| Use direct approximation rather than pruning (ablation) | AlexNet filter approximation · ImageNet ILSVRC-2012 learnable layers | accumulative energy | Mixed |
Closest papers by meaning
- SketchNet: Sketch Classification With Web Images (CVPR 2016 · similarity 0.89)
- Sketch-GNN: Scalable Graph Neural Networks with Sublinear Training Complexity (NeurIPS 2022 · similarity 0.89)
- From Hashing to CNNs: Training Binary Weight Networks via Hashing (AAAI 2018 · similarity 0.88)
- Sketch Down the FLOPs: Towards Efficient Networks for Human Sketch (CVPR 2025 · similarity 0.87)
- Convolutions and More as Einsum: A Tensor Network Perspective with Advances for Second-Order Methods (NeurIPS 2024 · similarity 0.86)
- PICNN: A Pathway towards Interpretable Convolutional Neural Networks (AAAI 2024 · 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.