Research map / Architectures and efficiency / Pruning and efficient tuning
PCONV: The Missing but Desirable Sparsity in DNN Weight Pruning for Real-Time Execution on Mobile Devices
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 |
|---|---|---|---|
| Apply SCP pattern pruning to each convolution kernel | VGG-16 · ImageNet | top-5 accuracy: 91.7\% to $92.5\% | Helped |
| Apply SCP pattern pruning to each convolution kernel | ResNet-50 · ImageNet | top-5 accuracy: 92.7\% to $93.0\% | Helped |
| Use compiler-assisted PCONV inference instead of TFLite | mobile CPU · VGG-16/ResNet-50/MobileNet-v2 | speedup: $9.4\times$ to $39.2\times$ | Helped |
| Use compiler-assisted PCONV inference instead of TVM | mobile GPU · VGG-16/ResNet-50/MobileNet-v2 | speedup: $2.5\times$ to $11.4\times$ | Helped |
| Increase the pattern count from 4 to 12 (ablation) | PCONV VGG-16 · ImageNet, CPU | inference time (ms): 52.7 -> 105.2 | Hurt |
| Increase the pattern count from 4 to 12 (ablation) | PCONV VGG-16 · ImageNet, GPU | inference time (ms): 19.1 -> 32.1 | Hurt |
| Increase the pattern count from 4 to 12 (ablation) | PCONV VGG-16 · ImageNet | top-5 accuracy (%): 91.5 -> 91.6 | No clear effect |
| Increase the pattern count from 4 to 12 (ablation) | PCONV VGG-16 · CIFAR-10, CPU | inference time (ms): 2.7 -> 3.1 | Hurt |
| Increase the pattern count from 4 to 12 (ablation) | PCONV VGG-16 · CIFAR-10 | top-1 accuracy (%): 93.8 -> 93.8 | No clear effect |
| Combine pattern pruning with connectivity pruning | PCONV · ImageNet and CIFAR-10 | accuracy at higher compression | Helped |
Closest papers by meaning
- Faster CNNs with Direct Sparse Convolutions and Guided Pruning (ICLR 2017 · similarity 0.90)
- AutoCompress: An Automatic DNN Structured Pruning Framework for Ultra-High Compression Rates (AAAI 2020 · similarity 0.90)
- RT3D: Achieving Real-Time Execution of 3D Convolutional Neural Networks on Mobile Devices (AAAI 2021 · similarity 0.90)
- SPDY: Accurate Pruning with Speedup Guarantees (ICML 2022 · similarity 0.89)
- Computation and Memory-Efficient Model Compression with Gradient Reweighting (NeurIPS 2025 · similarity 0.89)
- DPFPS: Dynamic and Progressive Filter Pruning for Compressing Convolutional Neural Networks from Scratch (AAAI 2021 · similarity 0.89)
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.