Focus of Attention in Reinforcement Learning

dc.creatorLi,Lihong
dc.creatorBulitko,Vadim
dc.creatorGreiner,Russell
dc.date2007
dc.date.accessioned2024-02-06T12:55:44Z
dc.date.available2024-02-06T12:55:44Z
dc.descriptionClassification-based reinforcement learning (RL) methods have recently been pro-posed as an alternative to the traditional value-function based methods. These methods use a classifier to represent a policy, where the input (features) to the classifier is the state and theoutput (class label) for that state is the desired action. The reinforcement-learning community knows that focusing on more important states can lead to improved performance. In this paper,we investigate the idea of focused learning in the context of classification-based RL. Specifically, we define a useful notation of state importance, which we use to prove rigorous bounds on policyloss. Furthermore, we show that a classification-based RL agent may behave arbitrarily poorly if it treats all states as equally important.
dc.formattext/html
dc.identifierhttps://doi.org/10.3217/jucs-013-09-1246
dc.identifierhttps://lib.jucs.org/article/28849/
dc.identifier.urihttps://openrepository.mephi.ru/handle/123456789/9472
dc.languageen
dc.publisherJournal of Universal Computer Science
dc.relationinfo:eu-repo/semantics/altIdentifier/eissn/0948-6968
dc.relationinfo:eu-repo/semantics/altIdentifier/pissn/0948-695X
dc.rightsinfo:eu-repo/semantics/openAccess
dc.rightsJ.UCS License
dc.sourceJUCS - Journal of Universal Computer Science 13(9): 1246-1269
dc.subjectreinforcement learning
dc.subjectfunction approximation
dc.subjectgeneralization
dc.subjectattention
dc.titleFocus of Attention in Reinforcement Learning
dc.typeResearch Article
Файлы
Коллекции