Research map / Architectures and efficiency / Pruning and efficient tuning
An Image Enhancing Pattern-based Sparsity for Real-time Inference on Mobile Devices
Xiaolong Ma, Wei Niu, Tianyun Zhang, Sijia Liu, Sheng Lin, Hongjia Li, Wujie Wen, Xiang Chen, Jian Tang, Kaisheng Ma, Bin Ren, Yanzhi Wang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Prune VGG-16 to Phase 2 pattern sparsity at 8.0x | VGG-16 · CIFAR-10 | accuracy (%): 93.5 -> 93.4 | Hurt |
| Apply pattern plus connectivity pruning at 8.0x to ResNet-18 | ResNet-18 · CIFAR-10 | accuracy (%): 94.0 -> 94.7 | Improved |
| Apply pattern plus connectivity pruning at 8.0x to ResNet-50 | ResNet-50 · CIFAR-10 | accuracy (%): 94.2 -> 95.2 | Improved |
| Raise ResNet-18 compression from 4.0x to 8.0x | ResNet-18 pattern pruning · ImageNet | top-1 accuracy (%): 69.6 -> 67.1 | Hurt |
| Raise VGG-16 compression from 8.0x to 12.0x | VGG-16 pattern pruning · ImageNet | top-1 accuracy (%): 74.4 -> 73.6 | Hurt |
| Raise ResNet-50 compression from 3.9x to 5.8x | ResNet-50 pattern pruning · ImageNet | top-1 accuracy (%): 75.9 -> 75.6 | Hurt |
| Raise VGG-16 compression from 8.0x to 19.7x (Phase 1) | VGG-16 pattern pruning · CIFAR-10 | accuracy (%): 93.4 -> 93.2 | Hurt |
| Use the designed 8-pattern library (Phase 2) with connectivity sparsity instead of other libraries (ablation) | pattern + connectivity pruning · ImageNet (ResNet-18, 6x) | accuracy | Improved |
| Use designed 8 patterns (Phase 2) on each convolution kernel (ablation) | pattern pruning · CIFAR-10 and ImageNet | accuracy | Improved |
| Pattern-prune VGG-16 with Phase 2 library | VGG-16 · ImageNet | visual quality | Improved |
Closest papers by meaning
- PCONV: The Missing but Desirable Sparsity in DNN Weight Pruning for Real-Time Execution on Mobile Devices (AAAI 2020 · similarity 0.92)
- GRIM: A General, Real-Time Deep Learning Inference Framework for Mobile Devices Based on Fine-Grained Structured Weight Sparsity (TPAMI 2021 · similarity 0.92)
- RT3D: Achieving Real-Time Execution of 3D Convolutional Neural Networks on Mobile Devices (AAAI 2021 · similarity 0.89)
- Faster CNNs with Direct Sparse Convolutions and Guided Pruning (ICLR 2017 · similarity 0.88)
- Scaling Laws for Sparsely-Connected Foundation Models (ICLR 2024 · similarity 0.85)
- Effective Interplay between Sparsity and Quantization: From Theory to Practice (ICLR 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.
