Research map / Continual learning / Domain and prompt-based
Incremental Multi-Domain Learning with Network Latent Tensor Factorization
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 |
|---|---|---|---|
| jointly tensorize identically structured CNN blocks and share a task-agnostic core | ResNet-26 multi-domain adaptation · Visual Decathlon | mean accuracy | Mixed |
| jointly tensorize identically structured CNN blocks and share a task-agnostic core | ResNet-26 multi-domain adaptation · Visual Decathlon | Decathlon score: 3744 -> 3585 | Hurt |
| use a shared core with task-specific factors rather than layer-wise adaptation | incremental multi-domain learning · Visual Decathlon | task-specific parameter count: 7.5× reduction | Helped |
| use ImageNet rather than CIFAR100 as the source pretraining dataset (ablation) | Ours · Aircraft | top-1 accuracy (%): 41.7 -> 55.6 | Helped |
| use ImageNet rather than CIFAR100 as the source pretraining dataset (ablation) | Ours · DTD | top-1 accuracy (%): 37.55 -> 52.2 | Helped |
| use ImageNet rather than CIFAR100 as the source pretraining dataset (ablation) | Ours · DPed | top-1 accuracy (%): 99.82 -> 99.67 | Hurt |
| use ImageNet rather than CIFAR100 as the source pretraining dataset (ablation) | Ours · GTSR | top-1 accuracy (%): 99.98 -> 99.96 | Hurt |
| apply low-rank constraints to the ImageNet network (ablation) | Ours · ImageNet | classification performance | Hurt |
| use orthogonality regularization λ=0.1 instead of λ=0.001 (ablation) | Ours · DTD | top-1 accuracy (%): 51.0% -> 52.2% | Helped |
| use orthogonality regularization λ=0.1 instead of λ=0.01 (ablation) | Ours · vgg-flowers | top-1 accuracy (%): 83.8% -> 80.9% | Hurt |
| fine-tune only factors and selected small components, keeping the shared core frozen | Ours · limited-data DPed, DTD, GTSR, and UCF | classification accuracy | Helped |
Closest papers by meaning
- Dual Consolidation for Pre-Trained Model-Based Domain-Incremental Learning (CVPR 2025 · similarity 0.87)
- Weight Diffusion for Future: Learn to Generalize in Non-Stationary Environments (NeurIPS 2024 · similarity 0.86)
- Advancing Cross-domain Discriminability in Continual Learning of Vision-Language Models (NeurIPS 2024 · similarity 0.86)
- Geometry-aware training of factorized layers in tensor Tucker format (NeurIPS 2024 · similarity 0.86)
- Factor Augmented Tensor-on-Tensor Neural Networks (AAAI 2025 · similarity 0.86)
- RECAST: Reparameterized, Compact weight Adaptation for Sequential Tasks (ICLR 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.