Research map / Architectures and efficiency / Pruning and efficient tuning
PDP: Parameter-free Differentiable Pruning is All You Need
Minsik Cho, Saurabh Adya, Devang Naik
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 the PDP soft mask rather than using the same training configuration without it (ablation) | PDP · ImageNet1k, ResNet50 | Top-1 accuracy (%): 73.1 -> 74.7 | Improved |
| Add the PDP soft mask rather than using the same training configuration without it (ablation) | PDP · ImageNet1k, MobileNet-v1 | Top-1 accuracy (%): 66.8 -> 68.2 | Improved |
| Train for more epochs to match ACDC GPU cost (PDP-base+) (ablation) | PDP-base · ImageNet1k, ResNet18, 85.5% sparsity | Top-1 accuracy (%): 69.0 -> 69.5 | Improved |
| Train for more epochs to match ACDC GPU cost (PDP-base+) (ablation) | PDP-base · ImageNet1k, ResNet50, 89.8% sparsity | Top-1 accuracy (%): 74.7 -> 75.3 | Improved |
| Use STR's per-layer sparsity allocation with PDP (PDP-str) (ablation) | PDP · ImageNet1k, ResNet18, 85.5% sparsity | Top-1 accuracy (%): 69.0 -> 68.6 | Hurt |
| Use OptG's per-layer sparsity allocation with PDP (PDP-optg) (ablation) | PDP · ImageNet1k, ResNet18, 85.5% sparsity | Top-1 accuracy (%): 69.0 -> 68.5 | Hurt |
| Vary the mask temperature τ (ablation) | PDP on MobileNet-v2 · ImageNet1k | Top-1 accuracy | Mixed |
| Use PDP rather than LNM for 1:4 structured pruning | ResNet18 · ImageNet1k, 1:4 pruning | Top-1 accuracy (%): 65.1 -> 68.7 | Improved |
| Use PDP rather than LNM for 2:4 structured pruning | ResNet50 · ImageNet1k, 2:4 pruning | Top-1 accuracy (%): 74.6 -> 75.9 | Improved |
| Apply PDP to channel pruning | ResNet50 · ImageNet1k, channel pruning | Top-1 accuracy (%): 75.3 -> 75.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ACDC — AC/DC: Alternating Compressed/DeCompressed Training of Deep Neural Networks (NeurIPS 2021) · better on 3 of 3 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
- DLP: Dynamic Layerwise Pruning in Large Language Models (ICML 2025 · similarity 0.89)
- OSSCAR: One-Shot Structured Pruning in Vision and Language Models with Combinatorial Optimization (ICML 2024 · similarity 0.88)
- SlimGPT: Layer-wise Structured Pruning for Large Language Models (NeurIPS 2024 · similarity 0.88)
- MULTIFLOW: Shifting Towards Task-Agnostic Vision-Language Pruning (CVPR 2024 · similarity 0.88)
- MDP: Multidimensional Vision Model Pruning with Latency Constraint (CVPR 2025 · similarity 0.87)
- Balanced Token Pruning: Accelerating Vision Language Models Beyond Local Optimization (NeurIPS 2025 · similarity 0.87)
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.
