Episode-Based Active Learning with Bayesian Neural Networks
Abstract
We investigate different strategies for active learning with Bayesian deep neural networks. We focus our analysis on scenarios where new, unlabeled data is obtained episodically, such as commonly encountered in mobile robotics applications. An evaluation of different strategies for acquisition, updating, and final training on the CIFAR-10 dataset shows that incremental network updates with final training on the accumulated acquisition set are essential for best performance, while limiting the amount of required human labeling labor.
Cite
Text
Dayoub et al. "Episode-Based Active Learning with Bayesian Neural Networks." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2017. doi:10.1109/CVPRW.2017.75Markdown
[Dayoub et al. "Episode-Based Active Learning with Bayesian Neural Networks." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2017.](https://mlanthology.org/cvprw/2017/dayoub2017cvprw-episodebased/) doi:10.1109/CVPRW.2017.75BibTeX
@inproceedings{dayoub2017cvprw-episodebased,
title = {{Episode-Based Active Learning with Bayesian Neural Networks}},
author = {Dayoub, Feras and Sünderhauf, Niko and Corke, Peter I.},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
year = {2017},
pages = {498-500},
doi = {10.1109/CVPRW.2017.75},
url = {https://mlanthology.org/cvprw/2017/dayoub2017cvprw-episodebased/}
}