Dynamic Neural Network for Multi-Task Learning Searching Across Diverse Network Topologies
Abstract
In this paper, we present a new MTL framework that searches for structures optimized for multiple tasks with diverse graph topologies and shares features among tasks. We design a restricted DAG-based central network with read-in/read-out layers to build topologically diverse task-adaptive structures while limiting search space and time. We search for a single optimized network that serves as multiple task adaptive sub-networks using our three-stage training process. To make the network compact and discretized, we propose a flow-based reduction algorithm and a squeeze loss used in the training process. We evaluate our optimized network on various public MTL datasets and show ours achieves state-of-the-art performance. An extensive ablation study experimentally validates the effectiveness of the sub-module and schemes in our framework.
Cite
Text
Choi and Im. "Dynamic Neural Network for Multi-Task Learning Searching Across Diverse Network Topologies." Conference on Computer Vision and Pattern Recognition, 2023. doi:10.1109/CVPR52729.2023.00368Markdown
[Choi and Im. "Dynamic Neural Network for Multi-Task Learning Searching Across Diverse Network Topologies." Conference on Computer Vision and Pattern Recognition, 2023.](https://mlanthology.org/cvpr/2023/choi2023cvpr-dynamic/) doi:10.1109/CVPR52729.2023.00368BibTeX
@inproceedings{choi2023cvpr-dynamic,
title = {{Dynamic Neural Network for Multi-Task Learning Searching Across Diverse Network Topologies}},
author = {Choi, Wonhyeok and Im, Sunghoon},
booktitle = {Conference on Computer Vision and Pattern Recognition},
year = {2023},
pages = {3779-3788},
doi = {10.1109/CVPR52729.2023.00368},
url = {https://mlanthology.org/cvpr/2023/choi2023cvpr-dynamic/}
}