Research map / Architectures and efficiency / Pruning and efficient tuning
Pruning Neural Networks via Coresets and Convex Geometry: Towards No Assumptions
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 |
|---|---|---|---|
| Use coreset pruning at 90% parameter pruning | LeNet-300-100 · MNIST | Pruned Err. (%): 2.03 -> 1.98 | Helped |
| Increase pruning ratio from 90 to 92.6 (ablation) | LeNet-300-100 with coreset pruning · MNIST | Pruned Err. (%): 1.98 -> 2.64 | Hurt |
| Increase pruning ratio from 92.6 to 94.6 (ablation) | LeNet-300-100 with coreset pruning · MNIST | Pruned Err. (%): 2.64 -> 3.51 | Hurt |
| Use coreset pruning at 90% parameter pruning | VGG-16 · CIFAR-10 | Pruned Err. (%): 8.16 -> 5.9 | Helped |
| Use coreset pruning at 88% filter pruning | VGG-19 · CIFAR-10 | Pruned Err. (%): 6.02 -> 5.85 | Helped |
| Increase pruning ratio from 88 to 91.28 (ablation) | VGG-19 with coreset pruning · CIFAR-10 | Pruned Err. (%): 5.85 -> 6.23 | Hurt |
| Use coreset pruning at 87% filter pruning | VGG-16 · CIFAR-10 | Pruned Err. (%): 7.17 -> 6.63 | Helped |
| Use coreset pruning at 79.53% filter pruning | VGG-16 · CIFAR-10 | Pruned Err. (%): 7.06 -> 6.3 | Helped |
| Use coreset pruning | ResNet56 · CIFAR-10 | Pruned Err. (%): 6.42 -> 7.26 | Hurt |
| Use coreset pruning at 62% parameter pruning | ResNet50 · ImageNet (ILSVRC-2012) | Pruned Err. (%): 25.11 -> 24.87 | Helped |
| Fine-tune the pruned model (ablation) | VGG-19 · CIFAR-10 | accuracy | Helped |
Closest papers by meaning
- Data-Independent Neural Pruning via Coresets (ICLR 2020 · similarity 0.95)
- “Learning-Compression” Algorithms for Neural Net Pruning (CVPR 2018 · similarity 0.91)
- Data-Dependent Coresets for Compressing Neural Networks with Applications to Generalization Bounds (ICLR 2019 · similarity 0.90)
- Computation and Memory-Efficient Model Compression with Gradient Reweighting (NeurIPS 2025 · similarity 0.90)
- How Sparse Can We Prune A Deep Network: A Fundamental Limit Perspective (NeurIPS 2024 · similarity 0.89)
- $\mathbb{D}^2$ Pruning: Message Passing for Balancing Diversity & Difficulty in Data Pruning (ICLR 2024 · 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.