Research map / Architectures and efficiency / Pruning and efficient tuning
On Compressing Deep Models by Low Rank and Sparse Decomposition
Check 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 low-rank and sparse decomposition together instead of pruning alone | AlexNet · ImageNet validation | compression rate: 9× -> 10× | Helped |
| Use low-rank and sparse decomposition together instead of pruning alone | VGG-16 · ImageNet validation | compression rate: 13× -> 15× | Helped |
| Use low-rank and sparse decomposition together instead of low-rank tensor decomposition | GoogLeNet · ImageNet validation | compression rate: 2.84× -> 4.5× | Helped |
| Add asymmetric feature-map reconstruction to weight decomposition (ablation) | VGG-16 convolutional layers before retraining · ImageNet validation | accuracy | Helped |
| Replace GreBdec with GoDec, omitting the asymmetric feature-map reconstruction objective (ablation) | VGG-16 convolutional layers before retraining · ImageNet validation | accuracy | Hurt |
| Compress GoogLeNet convolutional layers with kernel size larger than 1 at 5× | GoogLeNet · ImageNet testing | Top-5 accuracy: 3.7% | Hurt |
Closest papers by meaning
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · similarity 0.90)
- Compressing Neural Networks: Towards Determining the Optimal Layer-wise Decomposition (NeurIPS 2021 · similarity 0.90)
- Model Preserving Compression for Neural Networks (NeurIPS 2022 · similarity 0.90)
- Forget the Data and Fine-Tuning! Just Fold the Network to Compress (ICLR 2025 · similarity 0.88)
- MCNC: Manifold-Constrained Reparameterization for Neural Compression (ICLR 2025 · similarity 0.88)
- Effective Interplay between Sparsity and Quantization: From Theory to Practice (ICLR 2025 · similarity 0.88)
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.