Learning Object-Centered Autotelic Behaviors with Graph Neural Networks

Ahmed Akakzia, Olivier Sigaud
Proceedings of The 1st Conference on Lifelong Learning Agents, PMLR 199:351-365, 2022.

Abstract

Although humans live in an open-ended world and endlessly face new challenges, they do not have to learn from scratch each time they face the next one. Rather, they have access to a handful of previously learned skills, which they rapidly adapt to new situations. In artificial intelligence, autotelic agents—which are intrinsically motivated to represent and set their own goals—exhibit promising skill adaptation capabilities. However, these capabilities are highly constrained by their policy and goal space representations. In this paper, we propose to investigate the impact of these representations on the learning capabilities of autotelic agents. We study different implementations of autotelic agents using four types of Graph Neural Networks policy representations and two types of goal spaces, either geometric or predicate-based. We show that combining object-centered architectures that are expressive enough with semantic relational goals enables an efficient transfer between skills and promotes behavioral diversity. We also release our graph-based implementations to encourage further research in this direction.

Cite this Paper

BibTeX
@InProceedings{pmlr-v199-akakzia22a, title = {Learning Object-Centered Autotelic Behaviors with Graph Neural Networks}, author = {Akakzia, Ahmed and Sigaud, Olivier}, booktitle = {Proceedings of The 1st Conference on Lifelong Learning Agents}, pages = {351--365}, year = {2022}, editor = {Chandar, Sarath and Pascanu, Razvan and Precup, Doina}, volume = {199}, series = {Proceedings of Machine Learning Research}, month = {22--24 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v199/akakzia22a/akakzia22a.pdf}, url = {https://proceedings.mlr.press/v199/akakzia22a.html}, abstract = {Although humans live in an open-ended world and endlessly face new challenges, they do not have to learn from scratch each time they face the next one. Rather, they have access to a handful of previously learned skills, which they rapidly adapt to new situations. In artificial intelligence, autotelic agents—which are intrinsically motivated to represent and set their own goals—exhibit promising skill adaptation capabilities. However, these capabilities are highly constrained by their policy and goal space representations. In this paper, we propose to investigate the impact of these representations on the learning capabilities of autotelic agents. We study different implementations of autotelic agents using four types of Graph Neural Networks policy representations and two types of goal spaces, either geometric or predicate-based. We show that combining object-centered architectures that are expressive enough with semantic relational goals enables an efficient transfer between skills and promotes behavioral diversity. We also release our graph-based implementations to encourage further research in this direction.} }
Endnote
%0 Conference Paper %T Learning Object-Centered Autotelic Behaviors with Graph Neural Networks %A Ahmed Akakzia %A Olivier Sigaud %B Proceedings of The 1st Conference on Lifelong Learning Agents %C Proceedings of Machine Learning Research %D 2022 %E Sarath Chandar %E Razvan Pascanu %E Doina Precup %F pmlr-v199-akakzia22a %I PMLR %P 351--365 %U https://proceedings.mlr.press/v199/akakzia22a.html %V 199 %X Although humans live in an open-ended world and endlessly face new challenges, they do not have to learn from scratch each time they face the next one. Rather, they have access to a handful of previously learned skills, which they rapidly adapt to new situations. In artificial intelligence, autotelic agents—which are intrinsically motivated to represent and set their own goals—exhibit promising skill adaptation capabilities. However, these capabilities are highly constrained by their policy and goal space representations. In this paper, we propose to investigate the impact of these representations on the learning capabilities of autotelic agents. We study different implementations of autotelic agents using four types of Graph Neural Networks policy representations and two types of goal spaces, either geometric or predicate-based. We show that combining object-centered architectures that are expressive enough with semantic relational goals enables an efficient transfer between skills and promotes behavioral diversity. We also release our graph-based implementations to encourage further research in this direction.
APA
Akakzia, A. & Sigaud, O.. (2022). Learning Object-Centered Autotelic Behaviors with Graph Neural Networks. Proceedings of The 1st Conference on Lifelong Learning Agents, in Proceedings of Machine Learning Research 199:351-365 Available from https://proceedings.mlr.press/v199/akakzia22a.html.

Related Material