Research map / Architectures and efficiency / Neural architecture search
AdaXpert: Adapting Neural Architecture for Growing Data
Shuaicheng Niu, Jiaxiang Wu, Guanghui Xu, Yifan Zhang, Yong Guo, Peilin Zhao, Peng Wang, Mingkui Tan
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 |
|---|---|---|---|
| Adapt the previous architecture using data-distribution difference and prior architecture, rather than searching from scratch at each growth step | AdaXpert on ImageNet-100 · ImageNet-100, growing data snapshots | Acc.: 64.90 vs 64.80; 73.28 vs 73.46; 79.28 vs 78.88; 80.74 vs 80.62 | Mixed |
| Adapt the previous architecture instead of searching from scratch at each growth step | AdaXpert on ImageNet-100 · ImageNet-100, 40%, 80% and 100% data | Search Cost (GPU days): 1.0 vs 0.6; 1.3 vs 0.6; 1.5 vs 0.7 | Improved |
| Adjust the architecture for similar incoming data (ablation) | AdaXpert, base dataset plus small-difference data · ImageNet-100, similar-data condition | Acc.: 64.90 -> 65.04 (+0.14) | No clear effect |
| Adjust the architecture for different incoming data (ablation) | AdaXpert, base dataset plus large-difference data · ImageNet-100, different-data condition | Acc.: 72.54 -> 73.58 (+1.04) | Improved |
| Use the previous architecture and data difference to adapt the network | AdaXpert on ImageNet-1000 · ImageNet-1000, 1000 classes | Acc.: 78.13 vs 77.53 | Improved |
| Use AdaXpert rather than DEN | ImageNet-1000, 100 classes · ImageNet-1000, 100 classes | MAdds: 257 vs 515 | Improved |
Closest papers by meaning
- Adapting Neural Architectures Between Domains (NeurIPS 2020 · similarity 0.87)
- Rapid Network Adaptation: Learning to Adapt Neural Networks Using Test-Time Feedback (ICCV 2023 · similarity 0.86)
- Prodigy: An Expeditiously Adaptive Parameter-Free Learner (ICML 2024 · similarity 0.85)
- STAR: Synthesis of Tailored Architectures (ICLR 2025 · similarity 0.84)
- When Dynamic Data Selection Meets Data Augmentation: Achieving Enhanced Training Acceleration (ICML 2025 · similarity 0.84)
- SURGEON: Memory-Adaptive Fully Test-Time Adaptation via Dynamic Activation Sparsity (CVPR 2025 · similarity 0.84)
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.
