DANCE: Resource-Efficient Neural Architecture Search with Data-Aware and Continuous Adaptation

Wang, Maolin; Wei, Tianshuo; Zhang, Sheng; Guo, Ruocheng; Wang, Wangyu; Ye, Shanshan; Zou, Lixin; Wei, Xuetao; Zhao, Xiangyu

doi:10.24963/ijcai.2025/705

DANCE: Resource-Efficient Neural Architecture Search with Data-Aware and Continuous Adaptation

Maolin Wang, Tianshuo Wei, Sheng Zhang, Ruocheng Guo, Wangyu Wang, Shanshan Ye, Lixin Zou, Xuetao Wei, Xiangyu Zhao

Download BibTeX Cite

Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence. Main Track. Pages 6334-6342.

https://doi.org/10.24963/ijcai.2025/705

Keywords

Machine Learning: ML: Supervised Learning
Machine Learning: ML: Deep learning architectures

Abstract

Neural Architecture Search (NAS) has emerged as a powerful approach for automating neural network design. However, existing NAS methods face critical limitations in real-world deployments: architectures lack adaptability across scenarios, each deployment context requires costly separate searches, and performance consistency across diverse platforms remains challenging. We propose DANCE (Dynamic Architectures with Neural Continuous Evolution), which reformulates architecture search as a continuous evolution problem through learning distributions over architectural components. DANCE introduces three key innovations: a continuous architecture distribution enabling smooth adaptation, a unified architecture space with learned selection gates for efficient sampling, and a multi-stage training strategy for effective deployment optimization. Extensive experiments across five datasets demonstrate DANCE's effectiveness. Our method consistently outperforms state-of-the-art NAS approaches in terms of accuracy while significantly reducing search costs. Under varying computational constraints, DANCE maintains robust performance while smoothly adapting architectures to different hardware requirements. The code and appendix can be found at https://github.com/Applied-Machine-Learning-Lab/DANCE.

Maolin Wang, Tianshuo Wei, Sheng Zhang, Ruocheng Guo, Wangyu Wang, Shanshan Ye, Lixin Zou, Xuetao Wei, and Xiangyu Zhao. 2025. DANCE: Resource-Efficient Neural Architecture Search with Data-Aware and Continuous Adaptation. In Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-25. IJCAI, 6334–6342. https://doi.org/10.24963/ijcai.2025/705

@inproceedings{ijcai2025p705,
  title     = {DANCE: Resource-Efficient Neural Architecture Search with Data-Aware and Continuous Adaptation},
  author    = {Wang, Maolin and Wei, Tianshuo and Zhang, Sheng and Guo, Ruocheng and Wang, Wangyu and Ye, Shanshan and Zou, Lixin and Wei, Xuetao and Zhao, Xiangyu},
  booktitle = {Proceedings of the Thirty-Fourth International Joint Conference on
               Artificial Intelligence, {IJCAI-25}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {James Kwok},
  pages     = {6334--6342},
  year      = {2025},
  month     = {8},
  note      = {Main Track},
  doi       = {10.24963/ijcai.2025/705},
  url       = {https://doi.org/10.24963/ijcai.2025/705},
}

TY  - CPAPER
TI  - DANCE: Resource-Efficient Neural Architecture Search with Data-Aware and Continuous Adaptation
AU  - Wang, Maolin
AU  - Wei, Tianshuo
AU  - Zhang, Sheng
AU  - Guo, Ruocheng
AU  - Wang, Wangyu
AU  - Ye, Shanshan
AU  - Zou, Lixin
AU  - Wei, Xuetao
AU  - Zhao, Xiangyu
T2  - Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, IJCAI-25
PB  - International Joint Conferences on Artificial Intelligence Organization
PY  - 2025
SP  - 6334
EP  - 6342
DO  - 10.24963/ijcai.2025/705
UR  - https://doi.org/10.24963/ijcai.2025/705
ER  -