Efficient Reinforcement Learning using Gaussian Processes

File Description SizeFormat 
978-3-86644-569-7_pdfa.pdfPublished version5.51 MBAdobe PDFView/Open
Title: Efficient Reinforcement Learning using Gaussian Processes
Authors: Deisenroth, MP
Item Type: Book
Abstract: This book examines Gaussian processes (GPs) in model-based reinforcement learning (RL) and inference in nonlinear dynamic systems. First, we introduce PILCO, a fully Bayesian approach for efficient RL in continuous-valued state and action spaces when no expert knowledge is available. PILCO learns fast since it takes model uncertainties consistently into account during long-term planning and decision making. Thus, it reduces model bias, a common problem in model-based RL. Due to its generality and efficiency, PILCO is a conceptual and practical approach to jointly learning models and controllers fully automatically. Across all tasks, we report an unprecedented degree of automation and an unprecedented speed of learning. Second, we propose principled algorithms for robust filtering and smoothing in GP dynamic systems. Our methods are based on analytic moment matching and clearly advance state-of-the-art methods.
Editors: Hanebeck, UD
Issue Date: 30-Nov-2010
URI: http://hdl.handle.net/10044/1/12224
Publisher: KIT Scientific Publishing
Journal / Book Title: Karlsruhe Series on Intelligent Sensor-Actuator-Systems
Volume: 9
Copyright Statement: © 2009 The Author. This work is under the Creative Common license CC-BY-NC-ND. You are free to copy, distribute and transmit the work. You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work). You may not use this work for commercial purposes. You may not alter, transform, or build upon this work. http://creativecommons.org/licenses/by-nc-nd/3.0/de/
Notes: timestamp: 2010.11.17
Publisher URL: http://www.ksp.kit.edu/shop/isbn2shopid.php?isbn=978-3-86644-569-7
Appears in Collections:Computing



This item is licensed under a Creative Commons License Creative Commons