57,499 Hits in 6.4 sec

Partial Order Hierarchical Reinforcement Learning [chapter]

Bernhard Hengst
2008 Lecture Notes in Computer Science  
We go further and show how a problem can be automatically decomposed into a partial-order task-hierarchy, and solved using hierarchical reinforcement learning.  ...  In this paper the notion of a partial-order plan is extended to task-hierarchies. We introduce the concept of a partial-order taskhierarchy that decomposes a problem using multi-tasking actions.  ...  Acknowledgements NICTA is funded by the Australian Government as represented by the Department of Broadband, Communications and the Digital Economy and the Australian Research Council through the ICT Centre  ... 
doi:10.1007/978-3-540-89378-3_14 fatcat:bi6pxe4findg5oobpzqn4t6fha

State and Action Abstraction in the Development of Agent Controllers [chapter]

Brent E., Dean F.
2010 Autonomous Agents  
We would like to especially thank to John Antonio, Sesh Commuri, Andrew Fagg, and Amy McGovern for their input and insights.  ...  Acknowledgements We would like to thank all the members of the AI research group at the University of Oklahoma for their contributions.  ...  Unlike modular reinforcement learning, composite reinforcement learning does not attempt to learn policies for the primitive tasks simultaneously.  ... 
doi:10.5772/9655 fatcat:yzq6mndi6fhfjhlitmrjhrqxau

Object-Oriented Representation and Hierarchical Reinforcement Learning in Infinite Mario

M. Joshi, R. Khobragade, S. Sarda, U. Deshpande, S. Mohan
2012 2012 IEEE 24th International Conference on Tools with Artificial Intelligence  
In this work, we analyze and improve upon reinforcement learning techniques used to build agents that can learn to play Infinite Mario, an action game.  ...  We also extend the idea of hierarchical RL by designing a hierarchy in action selection using domain specific knowledge.  ...  In our solution, the levels in the hierarchy use an SMDP model for reinforcement learning to choose among temporally extended actions IV.  ... 
doi:10.1109/ictai.2012.152 dblp:conf/ictai/JoshiKSDM12 fatcat:dkj4fj6jonb57oxcsvtkwb67qy

TeXDYNA: Hierarchical Reinforcement Learning in Factored MDPs [chapter]

Olga Kozlova, Olivier Sigaud, Christophe Meyer
2010 Lecture Notes in Computer Science  
Reinforcement learning is one of the main adaptive mechanisms that is both well documented in animal behaviour and giving rise to computational studies in animats and robots.  ...  Learning and factorization techniques of Factored Reinforcement Learning.  ...  the overall state-action space into a set of smaller state-action spaces each of which can be factored.  ... 
doi:10.1007/978-3-642-15193-4_46 fatcat:5moziwlpdnantdq4dp5qim4k34

Hierarchical Reinforcement Learning with Hindsight [article]

Andrew Levy, Robert Platt, Kate Saenko
2019 arXiv   pre-print
We introduce a solution that enables agents to learn temporally extended actions at multiple levels of abstraction in a sample efficient and automated fashion.  ...  We show that our method significantly accelerates learning in a variety of discrete and continuous tasks.  ...  Acknowledgements This research was supported by NSF award IIS-1724237 and by DARPA.  ... 
arXiv:1805.08180v2 fatcat:2ijj7xoz65fplber4lxisk3onu

A neural model of hierarchical reinforcement learning

Daniel Rasmussen, Aaron Voelker, Chris Eliasmith, Gennady Cymbalyuk
2017 PLoS ONE  
We present the first model capable of performing hierarchical reinforcement learning in a general, neurally detailed implementation.  ...  We show that this model is able to learn a spatial pickup and delivery task more quickly than one without hierarchical abilities.  ...  Acknowledgements This work was supported by the Natural Sciences and Engineering Research Council of Canada, Canada Research Chairs, the Canadian Foundation for Innovation, and Ontario Innovation Trust  ... 
doi:10.1371/journal.pone.0180234 pmid:28683111 pmcid:PMC5500327 fatcat:3dzfenv2dfb3ngrnecm7q5zjma

Hierarchical Primitive Composition: Simultaneous Activation of Skills with Inconsistent Action Dimensions in Multiple Hierarchies [article]

Jeong-Hoon Lee, Jongeun Choi
2022 arXiv   pre-print
In this study, we further expand the discussion by incorporating simultaneous activation of the skills and structuring them into multiple hierarchies in a recursive fashion.  ...  Deep reinforcement learning has shown its effectiveness in various applications, providing a promising direction for solving tasks with high complexity.  ...  Top plot shows the effect of the simultaneous activation, plot in the middle shows the effect of the multi-level hierarchy, and the bottom plot shows the effect of our action composition. trainable auxiliary  ... 
arXiv:2110.01833v4 fatcat:ah3zmprekrf35bive6zatnrgxi

Bayesian exploration and interactive demonstration in continuous state MAXQ-learning

Kathrin Grave, Sven Behnke
2014 2014 IEEE International Conference on Robotics and Automation (ICRA)  
In this paper, we apply the MAXQ method for hierarchical reinforcement learning to continuous state spaces.  ...  To further reduce risk and to accelerate learning, we complement MAXQ with learning from demonstrations in an interactive way.  ...  At the same time, in practical applications, a reinforcement learning agent will only ever visit a small part of the state and action spaces.  ... 
doi:10.1109/icra.2014.6907337 dblp:conf/icra/GraveB14 fatcat:qs24z2ffgjbhdadasuyq7tiuaa

Hierarchical Learning in Stochastic Domains: Preliminary Results [chapter]

Leslie Pack Kaelbling
1993 Machine Learning Proceedings 1993  
This paper presents the HDG learning algorithm, which uses a hierarchical decomposition of the state space to make learning to achieve goals more efficient with a small penalty in path quality.  ...  The HDG algorithm, which is a descendent of Watkins' Q-learning algorithm, is described here and preliminary empirical results are presented.  ...  Acknowledgements This work was supported in part by a National Science Foundation National Young Investigator Award IRI-9257592 and in part by ONR Contract N00014-91-4052, ARPA Order 8225.  ... 
doi:10.1016/b978-1-55860-307-3.50028-9 dblp:conf/icml/Kaelbling93 fatcat:jhxt6fv4tzfrbkwsbkwvhcvmqi

Scalable reinforcement learning through hierarchical decompositions for weakly-coupled problems

Hazem Toutounji, Constantin A. Rothkopf, Jochen Triesch
2011 2011 IEEE International Conference on Development and Learning (ICDL)  
Reinforcement Learning, or Reward-Dependent Learning, has been very successful at describing how animals and humans adjust their actions so as to increase their gains and reduce their losses in a wide  ...  This suggests the existence of learning algorithms that are capable of taking advantage of the independencies present in the world and hence reducing the computational costs in terms of representations  ...  In a Parallel Schedule, the two levels of the hierarchy learn simultaneously.  ... 
doi:10.1109/devlrn.2011.6037351 dblp:conf/icdl-epirob/ToutounjiRT11 fatcat:nwemrichjffj5jn2idgguu3dmq

Efficient Skill Learning using Abstraction Selection

George Dimitri Konidaris, Andrew G. Barto
2009 International Joint Conference on Artificial Intelligence  
reinforcement learning domain.  ...  We show empirically that it can consistently select an appropriate abstraction using very little sample data, and that it significantly improves skill learning performance in a reasonably large real-valued  ...  Andrew Barto was supported by the Air Force Office of Scientific Research under grant FA9550-08-1-0418.  ... 
dblp:conf/ijcai/KonidarisB09 fatcat:bphg3b2igrgtjc6wciit23u3ki

Hierarchical Approaches [chapter]

Bernhard Hengst
2012 Adaptation, Learning, and Optimization  
Methods for reinforcement learning can be extended to work with abstract states and actions over a hierarchy of subtasks that decompose the original problem, potentially reducing its computational complexity  ...  This Chapter introduces hierarchical approaches to reinforcement learning that hold out the promise of reducing a reinforcement learning problems to a manageable size.  ...  Much of the literature in reinforcement learning involves one-dimensional actions. However, in many domains we wish to control several action variables simultaneously.  ... 
doi:10.1007/978-3-642-27645-3_9 fatcat:wofwukk2dbgdjn7x3dm3b4g3hq

Feudal Reinforcement Learning

Peter Dayan, Geoffrey E. Hinton
1992 Neural Information Processing Systems  
One way to speed up reinforcement learning is to enable learning to happen simultaneously at multiple resolutions in space and time.  ...  They simply learn to maximise their reinforcement in the context of the current command. We illustrate the system using a simple maze task ..  ...  This work was supported by SERC, the Howard Hughes Medical Institute and the Canadian Institute for Advanced Research (CIAR). GEH is the Noranda fellow of the CIAR.  ... 
dblp:conf/nips/DayanH92 fatcat:pw4pjqmoprhu3baer6axodjjza

Nonstrict Hierarchical Reinforcement Learning for Interactive Systems and Robots

Heriberto Cuayáhuitl, Ivana Kruijff-Korbayová, Nina Dethlefs
2014 ACM transactions on interactive intelligent systems (TiiS)  
To this end, each reinforcement learning agent in the hierarchy is extended with a subtask transition function and a dynamic state space to allow flexible switching between subdialogues.  ...  Conversational systems and robots that use reinforcement learning for policy optimization in large domains often face the problem of limited scalability.  ...  ACKNOWLEDGMENTS This work was supported by the European FP7 research projects ALIZ-E (ICT-248116), PARLANCE (287615), and SPACE-BOOK (270019).  ... 
doi:10.1145/2659003 fatcat:ns7ood3eyfatfoyqje5bcx4lpy

Latent Space Policies for Hierarchical Reinforcement Learning [article]

Tuomas Haarnoja, Kristian Hartikainen, Pieter Abbeel, Sergey Levine
2018 arXiv   pre-print
We address the problem of learning hierarchical deep neural network policies for reinforcement learning.  ...  task, but acquires a range of diverse strategies via a maximum entropy reinforcement learning objective.  ...  Acknowledgments We thank Aurick Zhou for producing some of the baseline results. This work was supported by Siemens and Berkeley DeepDrive.  ... 
arXiv:1804.02808v2 fatcat:p5dzbnrrfrhrxhm532jtbxvtfu
« Previous Showing results 1 — 15 out of 57,499 results