Testing the Limits of Data Efficiency in Experience Replay
Abstract
In Continual Learning rehearsal-based methods store a subset of observed data in a buffer for replay during training. The computational efficiency of these methods is tied to their data efficiency, i.e., the size of their buffer. In this work we expose a nuanced picture of rehearsal, underscoring the role of implicit biases on the road towards scalable CL.
Cite
Text
Meier et al. "Testing the Limits of Data Efficiency in Experience Replay." NeurIPS 2024 Workshops: Continual_FoMo, 2024.Markdown
[Meier et al. "Testing the Limits of Data Efficiency in Experience Replay." NeurIPS 2024 Workshops: Continual_FoMo, 2024.](https://mlanthology.org/neuripsw/2024/meier2024neuripsw-testing/)BibTeX
@inproceedings{meier2024neuripsw-testing,
title = {{Testing the Limits of Data Efficiency in Experience Replay}},
author = {Meier, Damiano and Lanzillotta, Giulia and Hofmann, Thomas},
booktitle = {NeurIPS 2024 Workshops: Continual_FoMo},
year = {2024},
url = {https://mlanthology.org/neuripsw/2024/meier2024neuripsw-testing/}
}