Meta-Learning by Adjusting Priors Based on Extended PAC-Bayes Theory

Ron Amit, Ron Meir
Proceedings of the 35th International Conference on Machine Learning, PMLR 80:205-214, 2018.

Abstract

In meta-learning an agent extracts knowledge from observed tasks, aiming to facilitate learning of novel future tasks. Under the assumption that future tasks are ‘related’ to previous tasks, accumulated knowledge should be learned in such a way that they capture the common structure across learned tasks, while allowing the learner sufficient flexibility to adapt to novel aspects of a new task. We present a framework for meta-learning that is based on generalization error bounds, allowing us to extend various PAC-Bayes bounds to meta-learning. Learning takes place through the construction of a distribution over hypotheses based on the observed tasks, and its utilization for learning a new task. Thus, prior knowledge is incorporated through setting an experience-dependent prior for novel tasks. We develop a gradient-based algorithm, and implement it for deep neural networks, based on minimizing an objective function derived from the bounds, and demonstrate its effectiveness numerically. In addition to establishing the improved performance available through meta-learning, we demonstrate the intuitive way by which prior information is manifested at different levels of the network.

Cite this Paper

BibTeX
@InProceedings{pmlr-v80-amit18a, title = {Meta-Learning by Adjusting Priors Based on Extended {PAC}-{B}ayes Theory}, author = {Amit, Ron and Meir, Ron}, booktitle = {Proceedings of the 35th International Conference on Machine Learning}, pages = {205--214}, year = {2018}, editor = {Dy, Jennifer and Krause, Andreas}, volume = {80}, series = {Proceedings of Machine Learning Research}, month = {10--15 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v80/amit18a/amit18a.pdf}, url = {https://proceedings.mlr.press/v80/amit18a.html}, abstract = {In meta-learning an agent extracts knowledge from observed tasks, aiming to facilitate learning of novel future tasks. Under the assumption that future tasks are ‘related’ to previous tasks, accumulated knowledge should be learned in such a way that they capture the common structure across learned tasks, while allowing the learner sufficient flexibility to adapt to novel aspects of a new task. We present a framework for meta-learning that is based on generalization error bounds, allowing us to extend various PAC-Bayes bounds to meta-learning. Learning takes place through the construction of a distribution over hypotheses based on the observed tasks, and its utilization for learning a new task. Thus, prior knowledge is incorporated through setting an experience-dependent prior for novel tasks. We develop a gradient-based algorithm, and implement it for deep neural networks, based on minimizing an objective function derived from the bounds, and demonstrate its effectiveness numerically. In addition to establishing the improved performance available through meta-learning, we demonstrate the intuitive way by which prior information is manifested at different levels of the network.} }
Endnote
%0 Conference Paper %T Meta-Learning by Adjusting Priors Based on Extended PAC-Bayes Theory %A Ron Amit %A Ron Meir %B Proceedings of the 35th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2018 %E Jennifer Dy %E Andreas Krause %F pmlr-v80-amit18a %I PMLR %P 205--214 %U https://proceedings.mlr.press/v80/amit18a.html %V 80 %X In meta-learning an agent extracts knowledge from observed tasks, aiming to facilitate learning of novel future tasks. Under the assumption that future tasks are ‘related’ to previous tasks, accumulated knowledge should be learned in such a way that they capture the common structure across learned tasks, while allowing the learner sufficient flexibility to adapt to novel aspects of a new task. We present a framework for meta-learning that is based on generalization error bounds, allowing us to extend various PAC-Bayes bounds to meta-learning. Learning takes place through the construction of a distribution over hypotheses based on the observed tasks, and its utilization for learning a new task. Thus, prior knowledge is incorporated through setting an experience-dependent prior for novel tasks. We develop a gradient-based algorithm, and implement it for deep neural networks, based on minimizing an objective function derived from the bounds, and demonstrate its effectiveness numerically. In addition to establishing the improved performance available through meta-learning, we demonstrate the intuitive way by which prior information is manifested at different levels of the network.
APA
Amit, R. & Meir, R.. (2018). Meta-Learning by Adjusting Priors Based on Extended PAC-Bayes Theory. Proceedings of the 35th International Conference on Machine Learning, in Proceedings of Machine Learning Research 80:205-214 Available from https://proceedings.mlr.press/v80/amit18a.html.

Related Material