REINFORCEMENT LEARNING IN NONSTATIONARY ENVIRONMENTS

منو

" REINFORCEMENT LEARNING IN NONSTATIONARY ENVIRONMENTS "

Document Type	:	Latin Dissertation
Language of Document	:	English
Record Number	:	151606
Doc. No	:	ET23398
Main Entry	:	PING-MAN CHOI
Title Proper	:	REINFORCEMENT LEARNING IN NONSTATIONARY ENVIRONMENTS
Note	:	This document is digital این مدرک بصورت الکترونیکی می باشد
Abstract	:	Learning to act optimally in the complex world has long been a major goal inartificial intelligence research. Reinforcement learning (RL) is an active researcharea that attempts to achieve this goal. In the past, studies on RL have beenfocused mainly on stationary environments, in which the underlying dynamics donot change over time. This assumption, however, is often unrealistic for real-worldtasks. Previous works on nonstationary RL tmically assume that nonstationaryenvironments change slowly enough for adaptation to take place. The dominatingapproach thus far is to employ online Learning techniques to give higher emphasison recent experience in order to cope with environmental changes. This onlinelearning approach is memoryless in the sense that even if the environment doesever revert to its previous dlvnamics,.....-....,....-...,..tested for theQ1 PC1 bus cardBoth these projects mere sofixare des elopment efforts tonards contributing to dlfferentaspects of Roboucs and lZ1echatronics projects m the Controls and Roboucs Group..
Subject	:	Electericl tess
	:	برق
electronic file name	:	TL46637.pdf
Title and statement of responsibility and	:	REINFORCEMENT LEARNING IN NONSTATIONARY ENVIRONMENTS [Thesis]