Research map / Architectures and efficiency / Pruning and efficient tuning
Disentangled Differentiable Network Pruning
Shangqian Gao, Feihu Huang, Yanfu Zhang, Heng Huang
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 |
|---|---|---|---|
| add disentangled width/importance pruning vs Gumbel-sigmoid differentiable pruning baseline (ablation) | DDNP · CIFAR-10 ResNet-56 | accuracy during pruning: advantage obvious at p = 0.35 and p = 0.5 | Improved |
| use soft top-k (smoothstep) instead of hard top-k (gamma = 0) (ablation) | DDNP · CIFAR-10 ResNet-56 | accuracy during pruning: gap around 5% at large pruning rate | Improved |
| use GRU+FC importance generator instead of FC only or per-channel scores (ablation) | DDNP · CIFAR-10 ResNet-56 | accuracy during pruning | Improved |
| use sigmoid output for gs instead of absolute or square function (ablation) | DDNP · CIFAR-10 ResNet-56 | accuracy during pruning | Improved |
| DDNP vs DCP pruning at similar FLOPs | DDNP · CIFAR-10 ResNet-56 | Delta-Acc: 0.20% better than DCP | Improved |
| DDNP vs GFS on MobileNetV2 | DDNP · ImageNet MobileNetV2 | Top-1: 0.60%/0.65% higher | Improved |
| DDNP vs GFS on MobileNetV3 small | DDNP · ImageNet MobileNetV3 small | Top-1: 1.23%/1.06% better | Improved |
| DDNP vs CC on ResNet-50 | DDNP · ImageNet ResNet-50 | Delta Top-1: improves GBN and CC by 0.43% and 0.32% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SCOP — SCOP: Scientific Control for Reliable Neural Network Pruning (NeurIPS 2020) · better on 3 of 3 shared results
- HRank — HRank: Filter Pruning Using High-Rank Feature Map (CVPR 2020) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers (ICLR 2018 · similarity 0.91)
- Network Pruning via Performance Maximization (CVPR 2021 · similarity 0.91)
- Operation-Aware Soft Channel Pruning using Differentiable Masks (ICML 2020 · similarity 0.91)
- A Probabilistic Approach to Neural Network Pruning (ICML 2021 · similarity 0.91)
- Revisiting Random Channel Pruning for Neural Network Compression (CVPR 2022 · similarity 0.91)
- DMCP: Differentiable Markov Channel Pruning for Neural Networks (CVPR 2020 · similarity 0.90)
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.
