Scholarworks Repository

Scalable distributed lifelong multi-task reinforcement learning -

Show simple item record

dc.contributor.author El Zini, Julia,
dc.date.accessioned 2018-10-11T11:43:20Z
dc.date.available 2018-10-11T11:43:20Z
dc.date.copyright 2020-02
dc.date.issued 2017
dc.date.submitted 2017
dc.identifier.other b21055762
dc.identifier.uri http://hdl.handle.net/10938/21495
dc.description Thesis. M.S. American University of Beirut. Department of Computer Science, 2017. T:6736$Advisor : Dr. Mohamad I. Jaber, Assitant Professor, Computer Science ; Members of Committee : Dr. Mariette Awad, Associate Professor, Electrical and Computer Engineering ; Dr. Wassim El Hajj, Associate Professor, Computer Science.
dc.description Includes bibliographical references (leaves 69-77)
dc.description.abstract Multi-task reinforcement learning (MTRL) suffers from scalability issues when the number of tasks or trajectories per task grows-large. One of the main reasons behind this limitation is the reliance on centralized solutions. Recent methods exploited the connection between MTRL and general consensus to propose scalable solutions with linear convergence guarantees. In this work, we improve over state-of-the-art by presenting a distributed solver for MTRL with quadratic convergence guarantees. Our algorithm exploits a novel connection between MTRL and Laplacian-based general consensus that leads to an efficient solver. We further extend our work to the lifelong settings where we propose the first distributed lifelong MTRL solver who exhibits vanishing regret. We analyze both the theoretical and empirical properties of our method. In set of extensive experiments, we also show that the novel algorithm outperforms state-of-the-art on a variety of dynamical systems, including a simulated humanoid robot.
dc.format.extent 1 online resource (ix, 77 leaves) ; illustrations
dc.language.iso eng
dc.subject.classification T:006736
dc.subject.lcsh Reinforcement learning.$Mathematical optimization.$Distributed artificial intelligence.$Multitasking (Computer science)
dc.title Scalable distributed lifelong multi-task reinforcement learning -
dc.type Thesis
dc.contributor.department Faculty of Arts and Sciences.$Department of Computer Science.
dc.subject.classificationsource AUBNO
dc.contributor.institution American University of Beirut.


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search Scholarworks


Browse

My Account