Publications• Sorted by Date • Classified by Publication Type • Classified by Research Category • Influence-Based Abstraction in Deep Reinforcement LearningMiguel Suau de Castro, Elena Congeduti, Rolf A.N. Starre, Aleksander Czechowski, and Frans A. Oliehoek. Influence-Based Abstraction in Deep Reinforcement Learning. In Proceedings of the AAMAS Workshop on Adaptive Learning Agents (ALA), May 2019. DownloadAbstractReal-world systems are typically extremely complex, consisting of thousands, oreven millions of state variables. Unfortunately, applying reinforcementlearning algorithms to handle complex tasks becomes more and morechallenging as the number of state variables increases. In this paper, webuild on the concept of <i>influence-based abstraction</i> which tries to tacklesuch scalability issues by decomposing large systems into small regions. We explore this method in the context of deep reinforcement learning, showing that by keeping track of a small set of variables in the history of previous actions and observations we can learn policies that can effectively control a local region in the global system. BibTeX Entry@inproceedings{Suau19ALA, title = {Influence-Based Abstraction in Deep Reinforcement Learning}, author = {Suau de Castro, Miguel and Elena Congeduti and Rolf A.N. Starre and Aleksander Czechowski and Frans A. Oliehoek}, booktitle = ALA19, year = 2019, month = may, url = {https://ala2019.vub.ac.be/papers/ALA2019_paper_35.pdf}, keywords = {refereed, workshop}, abstract={ Real-world systems are typically extremely complex, consisting of thousands, or even millions of state variables. Unfortunately, applying reinforcement learning algorithms to handle complex tasks becomes more and more challenging as the number of state variables increases. In this paper, we build on the concept of \emph{influence-based abstraction} which tries to tackle such scalability issues by decomposing large systems into small regions. We explore this method in the context of deep reinforcement learning, showing that by keeping track of a small set of variables in the history of previous actions and observations we can learn policies that can effectively control a local region in the global system. } }
Generated by
bib2html.pl
(written by Patrick Riley) on
Tue Dec 10, 2024 18:28:47 UTC |