東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Connectionist reinforcement learning...

Hougen, Dean Frederick.

FindBook

Google Book

Amazon

博客來

Connectionist reinforcement learning for control of robotic systems.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Connectionist reinforcement learning for control of robotic systems./
作者:	Hougen, Dean Frederick.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 1998,
面頁冊數:	191 p.
附註:	Source: Dissertations Abstracts International, Volume: 60-06, Section: B.
Contained By:	Dissertations Abstracts International60-06B.
標題:	Computer science. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=9907501
ISBN:	9780599052994

Connectionist reinforcement learning for control of robotic systems.
Hougen, Dean Frederick.

Connectionist reinforcement learning for control of robotic systems. - Ann Arbor : ProQuest Dissertations & Theses, 1998 - 191 p.

Source: Dissertations Abstracts International, Volume: 60-06, Section: B.

Thesis (Ph.D.)--University of Minnesota, 1998.

This item must not be sold to any third party vendors.

Reinforcement learning for control may provide intelligent robotic agents with a means of improving their performance through interaction with the environment. Unfortunately, existing reinforcement-learning schemes require great computational power, a large memory, and/or thousands of training trials to achieve good performance. These resource requirements make implementation of reinforcement learning on real robotic systems impractical. This thesis presents a system for reinforcement learning for control that uses laterally connected artificial neural networks to improve the efficiency of a basic reinforcement-learning mechanism. This system uses a Kohonen-type Self-Organizing Map to efficiently partition the input space of continuous sensor data. It is capable of rapid learning of output responses in temporal domains through the use of eligibility traces and data sharing within topologically defined neighborhoods. The system performance is demonstrated through extensive simulation results and implementation on a real mini-robot with low computing power and little memory. The system learns well despite noisy, low-grain sensory input and uncertain interactions between motor commands and effects in the world. The truck-backing task is used as an example of a difficult credit-assignment problem.

ISBN: 9780599052994Subjects--Topical Terms:

523869
Computer science.

Connectionist reinforcement learning for control of robotic systems.
LDR:02441nmm a2200337 4500 001 2206914
005 20190906083326.5
008 201008s1998 ||||||||||||||||| ||eng d
020 $a 9780599052994
035 $a (MiAaPQ)AAI9907501
035 $a AAI9907501
040 $a MiAaPQ $c MiAaPQ
100 1 $a Hougen, Dean Frederick. $3 3433839
245 1 0 $a Connectionist reinforcement learning for control of robotic systems.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 1998
300 $a 191 p.
500 $a Source: Dissertations Abstracts International, Volume: 60-06, Section: B.
500 $a Publisher info.: Dissertation/Thesis.
500 $a Slagle, James R.;Gini, Maria L.
502 $a Thesis (Ph.D.)--University of Minnesota, 1998.
506 $a This item must not be sold to any third party vendors.
506 $a This item must not be added to any third party search indexes.
520 $a Reinforcement learning for control may provide intelligent robotic agents with a means of improving their performance through interaction with the environment. Unfortunately, existing reinforcement-learning schemes require great computational power, a large memory, and/or thousands of training trials to achieve good performance. These resource requirements make implementation of reinforcement learning on real robotic systems impractical. This thesis presents a system for reinforcement learning for control that uses laterally connected artificial neural networks to improve the efficiency of a basic reinforcement-learning mechanism. This system uses a Kohonen-type Self-Organizing Map to efficiently partition the input space of continuous sensor data. It is capable of rapid learning of output responses in temporal domains through the use of eligibility traces and data sharing within topologically defined neighborhoods. The system performance is demonstrated through extensive simulation results and implementation on a real mini-robot with low computing power and little memory. The system learns well despite noisy, low-grain sensory input and uncertain interactions between motor commands and effects in the world. The truck-backing task is used as an example of a difficult credit-assignment problem.
590 $a School code: 0130.
650 4 $a Computer science. $3 523869
650 4 $a Artificial intelligence. $3 516317
650 4 $a Systems design. $3 3433840
690 $a 0984
690 $a 0800
690 $a 0790
710 2 $a University of Minnesota. $3 676231
773 0 $t Dissertations Abstracts International $g 60-06B.
790 $a 0130
791 $a Ph.D.
792 $a 1998
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=9907501