Transcription of Trust Region Policy Optimization
{{id}} {{{paragraph}}}
Example: marketing
Trust Region Policy Optimization
Trust Region Policy OptimizationJohn of California, Berkeley, Department of Electrical Engineering and Computer SciencesAbstractIn this article, we describe a method for optimiz-ing control policies, with guaranteed monotonicimprovement. By making several approxima-tions to the theoretically-justified scheme, we de-velop a practical algorithm, called Trust RegionPolicy Optimization (TRPO). This algorithm iseffective for optimizing large nonlinear poli-cies such as neural networks. Our experimentsdemonstrate its robust performance on a wide va-riety of tasks: learning simulated robotic swim-ming, hopping, and walking gaits; and playingAtari games using images of the screen as its approximations that deviate from thetheory, TRPO tends to give monotonic improve-ment, with little tuning of algorithms for Policy Optimization can be classifiedinto three broad categories: Policy iteration methods, whichalternate between estimating the value function under thecurrent Policy and improving the Policy (Bertsekas, 2005); Policy gradient methods, which use an estimator of the gra-dient of the expected cost obtained from sample trajec-tories (Peters & Schaal, 2008a) (and which, as we laterdiscuss, have a close connection to Policy iteration); andderiva
Learning, Lille, France, 2015. JMLR: W&CP volume 37. Copy-right 2015 by the author(s). namic programming (ADP) methods, stochastic optimiza-tion methods are difficult to beat on this task (Gabillon et al., 2013). For continuous control problems, methods like CMA have been successful at learning control poli-
Loading..
Tags:
Information
Domain:
Source:
Link to this page:
Please notify us if you found a problem with this document: