Treatment Effect Estimation to Guide Model Optimization in Continual Learning

Jonas Seng, Florian Peter Busch, Matej Zečević, Moritz Willig
Proceedings of The First AAAI Bridge Program on Continual Causality, PMLR 208:38-44, 2023.

Abstract

Continual Learning systems are faced with a potentially large numbers of tasks to be learned while the models employed have only limited capacity available, which makes it potentially impossible to learn all required tasks within a single model. In order to detect on when a model might break we propose to use treatment effect estimation techniques to estimate the effect of training a model on a new task w.r.t. some suitable performance measure.

Cite this Paper

BibTeX
@InProceedings{pmlr-v208-seng23a, title = {Treatment Effect Estimation to Guide Model Optimization in Continual Learning}, author = {Seng, Jonas and Busch, Florian Peter and Ze\v{c}evi\'c, Matej and Willig, Moritz}, booktitle = {Proceedings of The First AAAI Bridge Program on Continual Causality}, pages = {38--44}, year = {2023}, editor = {Mundt, Martin and Cooper, Keiland W. and Dhami, Devendra Singh and Ribeiro, Adéle and Smith, James Seale and Bellot, Alexis and Hayes, Tyler}, volume = {208}, series = {Proceedings of Machine Learning Research}, month = {07--08 Feb}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v208/seng23a/seng23a.pdf}, url = {https://proceedings.mlr.press/v208/seng23a.html}, abstract = {Continual Learning systems are faced with a potentially large numbers of tasks to be learned while the models employed have only limited capacity available, which makes it potentially impossible to learn all required tasks within a single model. In order to detect on when a model might break we propose to use treatment effect estimation techniques to estimate the effect of training a model on a new task w.r.t. some suitable performance measure.} }
Endnote
%0 Conference Paper %T Treatment Effect Estimation to Guide Model Optimization in Continual Learning %A Jonas Seng %A Florian Peter Busch %A Matej Zečević %A Moritz Willig %B Proceedings of The First AAAI Bridge Program on Continual Causality %C Proceedings of Machine Learning Research %D 2023 %E Martin Mundt %E Keiland W. Cooper %E Devendra Singh Dhami %E Adéle Ribeiro %E James Seale Smith %E Alexis Bellot %E Tyler Hayes %F pmlr-v208-seng23a %I PMLR %P 38--44 %U https://proceedings.mlr.press/v208/seng23a.html %V 208 %X Continual Learning systems are faced with a potentially large numbers of tasks to be learned while the models employed have only limited capacity available, which makes it potentially impossible to learn all required tasks within a single model. In order to detect on when a model might break we propose to use treatment effect estimation techniques to estimate the effect of training a model on a new task w.r.t. some suitable performance measure.
APA
Seng, J., Busch, F.P., Zečević, M. & Willig, M.. (2023). Treatment Effect Estimation to Guide Model Optimization in Continual Learning. Proceedings of The First AAAI Bridge Program on Continual Causality, in Proceedings of Machine Learning Research 208:38-44 Available from https://proceedings.mlr.press/v208/seng23a.html.

Related Material