Research map / Architectures and efficiency / Pruning and efficient tuning
Recall Distortion in Neural Network Pruning and the Undecayed Pruning Algorithm
Aidan Good, Jacky Lin, Xin Yu, Hannah Sieg, Mikey Fergurson, Shandian Zhe, Jerzy Wieczorek, Thiago Serra
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 |
|---|---|---|---|
| Use Undecayed Pruning instead of Magnitude Pruning | CIFAR-10 models · CIFAR-10, ResNet-56 | intensification | Improved |
| Use Undecayed Pruning instead of Gradient Pruning | CIFAR-10 models · CIFAR-10, ResNet-56 | intensification | Improved |
| Increase pruning ratio | Magnitude Pruning and Undecayed Pruning · CIFAR-10, CIFAR-100, MNIST, and Fashion-MNIST | intensification | Hurt |
| Use moderate rather than high pruning ratios | Pruned neural networks · Classification on balanced datasets | intensification | Improved |
| Use moderate pruning | Pruned neural networks · Classification on balanced datasets | accuracy | Improved |
| Use Random Pruning instead of Magnitude, Gradient, or Undecayed Pruning | CIFAR-10, ResNet-56 · CIFAR-10 | intensification | Improved |
| Increase model size | Magnitude- or Undecayed-Pruned networks · CIFAR-10 | intensification | Improved |
| Increase dataset/task complexity | Magnitude-Pruned networks · MNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100 | intensification | Hurt |
| Use Lottery Ticket Hypothesis pruning at higher pruning ratios | ResNet-56 · CIFAR-10 | intensification: 0.986 -> 1.244 | Hurt |
| Use CHIP at a higher pruning ratio | ResNet-56 · CIFAR-10 | intensification: 0.985 -> 1.494 | Hurt |
Closest papers by meaning
- Pruning Neural Networks at Initialization: Why Are We Missing the Mark? (ICLR 2021 · similarity 0.89)
- A Three-regime Model of Network Pruning (ICML 2023 · similarity 0.88)
- The Generalization-Stability Tradeoff In Neural Network Pruning (NeurIPS 2020 · similarity 0.88)
- Random Pruning Over-parameterized Neural Networks Can Improve Generalization: A Training Dynamics Analysis (JMLR 2025 · similarity 0.88)
- Distilling the Knowledge in Data Pruning (ICML 2025 · similarity 0.87)
- Network Pruning That Matters: A Case Study on Retraining Variants (ICLR 2021 · 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.
