Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain-Adaptive Deep Network Compression
Marc Masana, Joost van de Weijer, Luis Herranz, Andrew D. Bagdanov, Jose M. Alvarez
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 activation-aware DALR instead of weight-only truncated SVD | fc6 low-rank compression · CUB-200 Birds (k=32) | accuracy: 16.83 -> 48.81 | Improved |
| use DALR instead of truncated SVD | fc6 low-rank compression · Oxford 102 Flowers (k=32) | accuracy: 14.00 -> 72.17 | Improved |
| use DALR instead of truncated SVD | fc6 low-rank compression · Stanford 40 Actions (k=32) | accuracy: 38.18 -> 64.70 | Improved |
| use DALR instead of truncated SVD | fc7 low-rank compression · Oxford 102 Flowers (k=32) | accuracy: 58.14 -> 72.32 | Improved |
| use DALR instead of truncated SVD on the source domain | fc6 low-rank compression · ImageNet val (k=32) | top-1 error: 79.44 -> 66.43 | Improved |
| use DALR instead of truncated SVD on the source domain | fc7 low-rank compression · ImageNet val (k=32) | top-1 error: 57.07 -> 56.32 | No clear effect |
| add bias compensation to truncated SVD (ablation) | fc6 low-rank compression · CUB-200 Birds (k=32) | accuracy: 16.83 -> 27.91 | Improved |
| add bias compensation to truncated SVD (ablation) | fc6 low-rank compression · Stanford 40 Actions (k=32) | accuracy: 38.18 -> 46.76 | Improved |
| add bias compensation to truncated SVD (ablation) | fc7 low-rank compression · Oxford 102 Flowers (k=32) | accuracy: 58.14 -> 57.07 | Hurt |
| add bias compensation to truncated SVD (ablation) | fc7 low-rank compression · ImageNet val (k=32) | top-1 error: 57.07 -> 55.57 | Improved |
| add bias compensation on top of DALR (ablation) | DALR · fine-tuned VGG19 compression | accuracy | No clear effect |
| replace truncated SVD with activation-based pruning (mean) | fc6 compression · CUB-200 Birds (k=1024) | accuracy: 55.25 -> 50.41 | Hurt |
Closest papers by meaning
- Compression-aware Training of Deep Networks (NeurIPS 2017 · similarity 0.90)
- Model Preserving Compression for Neural Networks (NeurIPS 2022 · similarity 0.89)
- Few Sample Knowledge Distillation for Efficient Network Compression (CVPR 2020 · similarity 0.89)
- Low-Rank Compression of Neural Nets: Learning the Rank of Each Layer (CVPR 2020 · similarity 0.88)
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · similarity 0.87)
- Auto-Compressing Networks (NeurIPS 2025 · similarity 0.86)
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.
