BROWSE

ITEM VIEW & DOWNLOAD

Deep Reinforcement Learning in Multi-End Games

Cited 0 times inthomson ciCited 0 times inthomson ci
Title
Deep Reinforcement Learning in Multi-End Games
Other Titles
멀티 엔드 게임에서의 깊은 강화학습
Author
Kim, Sol A
Advisor
Kim, Kwang In
Keywords
강화학습; 커널회귀; 몬테카를로트리탐색; 컬링; 딥러닝
Issue Date
2020-02
Publisher
Graduate School of UNIST
Abstract
Recently deep reinforcement learning (DRL) algorithms show super human performances in the simulated game domains. In practical points, the sample efficiency is also one of the most important measures to determine the performance of a model. Especially for the environment of large search spaces (e.g. continuous action space), it is very critical condition to achieve the state-of-the-art performance. In this thesis, we design a model to be applicable to multi-end games in continuous space with high sample efficiency. A multi-end game has several sub-games which are independent each other but affect the result of the game by some rules of its domain. We verify the algorithm in the environment of simulated curling.
Description
Department of Computer Science and Engineering
URI
Go to Link
Appears in Collections:
EE_Theses_Master
Files in This Item:
Deep Reinforcement Learning in Multi-End Games.pdf Download

find_unist can give you direct access to the published full text of this article. (UNISTARs only)

Show full item record

qrcode

  • mendeley

    citeulike

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

MENU