Fast inference and transfer of compositional task structures for few-shot task generalization

Sungryull Sohn, Hyunjae Woo, Jongwook Choi, Lyubing Qiang, Izzeddin Gur, Aleksandra Faust, Honglak Lee
Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, PMLR 180:1857-1865, 2022.

Abstract

We tackle real-world problems with complex structures beyond the pixel-based game or simulator. We formulate it as a few-shot reinforcement learning problem where a task is characterized by a subtask graph that defines a set of subtasks and their dependencies that are unknown to the agent. Different from the previous meta-RL methods trying to directly infer the unstructured task embedding, our multi-task subtask graph inferencer (MTSGI) first infers the common high-level task structure in terms of the subtask graph from the training tasks, and use it as a prior to improve the task inference in testing. Our experiment results on 2D grid-world and complex web navigation domains show that the proposed method can learn and leverage the common underlying structure of the tasks for faster adaptation to the unseen tasks than various existing algorithms such as meta reinforcement learning, hierarchical reinforcement learning, and other heuristic agents.

Cite this Paper

BibTeX
@InProceedings{pmlr-v180-sohn22a, title = {Fast inference and transfer of compositional task structures for few-shot task generalization}, author = {Sohn, Sungryull and Woo, Hyunjae and Choi, Jongwook and Qiang, Lyubing and Gur, Izzeddin and Faust, Aleksandra and Lee, Honglak}, booktitle = {Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence}, pages = {1857--1865}, year = {2022}, editor = {Cussens, James and Zhang, Kun}, volume = {180}, series = {Proceedings of Machine Learning Research}, month = {01--05 Aug}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v180/sohn22a/sohn22a.pdf}, url = {https://proceedings.mlr.press/v180/sohn22a.html}, abstract = {We tackle real-world problems with complex structures beyond the pixel-based game or simulator. We formulate it as a few-shot reinforcement learning problem where a task is characterized by a subtask graph that defines a set of subtasks and their dependencies that are unknown to the agent. Different from the previous meta-RL methods trying to directly infer the unstructured task embedding, our multi-task subtask graph inferencer (MTSGI) first infers the common high-level task structure in terms of the subtask graph from the training tasks, and use it as a prior to improve the task inference in testing. Our experiment results on 2D grid-world and complex web navigation domains show that the proposed method can learn and leverage the common underlying structure of the tasks for faster adaptation to the unseen tasks than various existing algorithms such as meta reinforcement learning, hierarchical reinforcement learning, and other heuristic agents.} }
Endnote
%0 Conference Paper %T Fast inference and transfer of compositional task structures for few-shot task generalization %A Sungryull Sohn %A Hyunjae Woo %A Jongwook Choi %A Lyubing Qiang %A Izzeddin Gur %A Aleksandra Faust %A Honglak Lee %B Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2022 %E James Cussens %E Kun Zhang %F pmlr-v180-sohn22a %I PMLR %P 1857--1865 %U https://proceedings.mlr.press/v180/sohn22a.html %V 180 %X We tackle real-world problems with complex structures beyond the pixel-based game or simulator. We formulate it as a few-shot reinforcement learning problem where a task is characterized by a subtask graph that defines a set of subtasks and their dependencies that are unknown to the agent. Different from the previous meta-RL methods trying to directly infer the unstructured task embedding, our multi-task subtask graph inferencer (MTSGI) first infers the common high-level task structure in terms of the subtask graph from the training tasks, and use it as a prior to improve the task inference in testing. Our experiment results on 2D grid-world and complex web navigation domains show that the proposed method can learn and leverage the common underlying structure of the tasks for faster adaptation to the unseen tasks than various existing algorithms such as meta reinforcement learning, hierarchical reinforcement learning, and other heuristic agents.
APA
Sohn, S., Woo, H., Choi, J., Qiang, L., Gur, I., Faust, A. & Lee, H.. (2022). Fast inference and transfer of compositional task structures for few-shot task generalization. Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 180:1857-1865 Available from https://proceedings.mlr.press/v180/sohn22a.html.

Related Material