New Metrics and Experimental Paradigms for Continual Learning

Tyler L. Hayes, Ronald Kemker, Nathan D. Cahill, Christopher Kanan

CVPRW 2018 pp. 2031-2034

doi:10.1109/CVPRW.2018.00273 /cvprw/2018/hayes2018cvprw-new/

Abstract

In order for a robotic agent to learn successfully in an uncontrolled environment, it must be able to immediately alter its behavior. Deep neural networks are the dominant approach for classification tasks in computer vision, but typical algorithms and architectures are incapable of immediately learning new tasks without catastrophically forgetting previously acquired knowledge. There has been renewed interest in solving this problem, but there are limitations to existing solutions, including poor performance compared to offline models, large memory footprints, and learning slowly. In this Abstract, we formalize the continual learning paradigm and propose new benchmarks for assessing continual learning agents.

CVPRW Semantic Scholar

Cite

Text

Hayes et al. "New Metrics and Experimental Paradigms for Continual Learning." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2018. doi:10.1109/CVPRW.2018.00273

Markdown

[Hayes et al. "New Metrics and Experimental Paradigms for Continual Learning." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2018.](https://mlanthology.org/cvprw/2018/hayes2018cvprw-new/) doi:10.1109/CVPRW.2018.00273

BibTeX

@inproceedings{hayes2018cvprw-new,
  title     = {{New Metrics and Experimental Paradigms for Continual Learning}},
  author    = {Hayes, Tyler L. and Kemker, Ronald and Cahill, Nathan D. and Kanan, Christopher},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year      = {2018},
  pages     = {2031-2034},
  doi       = {10.1109/CVPRW.2018.00273},
  url       = {https://mlanthology.org/cvprw/2018/hayes2018cvprw-new/}
}