 |
English
|
正體中文
|
简体中文
|
Post-Print筆數 : 27 |
Items with full text/Total items : 110936/141856 (78%)
Visitors : 47735069
Online Users : 1056
|
|
|
Loading...
|
Please use this identifier to cite or link to this item:
https://nccur.lib.nccu.edu.tw/handle/140.119/53373
|
| Title: | 吃角子老虎問題之最佳貝氏策略 |
| Other Titles: | Optimal Bayesian Strategies for Bandit Problems |
| Authors: | 洪英超 |
| Contributors: | 國立政治大學統計學系
行政院國家科學委員會 |
| Keywords: | 多拉桿吃角子老虎問題;貝氏策略
Multi-armed bandit problem;Bayesian strategy |
| Date: | 2010 |
| Issue Date: | 2012-08-30 09:58:31 (UTC+8) |
| Abstract: | 多拉桿吃角子老虎問題(multi-armed bandit problem)可以應用在許多領域如臨床試驗,線 上工業實驗(on-line industrial experimentations),可調性網路路由（adaptive network routing)等. 本計畫將以貝氏的角度探討“無窮多拉桿之吃角子老虎問題“.我們假設未 知的白努利參數為相互獨立且來自同一個機率分配F,而我們的目的是找出一如何選擇 拉桿的策略使得長時間操作下的失敗率為最低. 在本計畫的第一部份,我們假設F為一 任意但已知的機率分配.接著介紹1996年由Berry等人提出的三種策略,並証明當試驗次 數趨近無窮大時,此三種策略皆可以使長時間操作下的失敗率為最低.此外,我們也利用 電腦模擬來比較此三種策略的實際表現. 在本計畫的第二部份,我們假設F為一未知的 機率分配.在此假設下,我們提出一個新的策略叫做” empirical non-recalling m-run策略”, 並証明此策略亦為一近似最佳策略. 此外,我們也將利用電腦模擬與Herschkorn等人於 1995年提出的二個策略進行比較.
Multi-armed bandit problems have a wide area of applications such as clinical trials, on-line industrial experimentations, adaptive network routing, etc. In this study, we examine the bandit problem with infinitely many arms from a Bayesian perspective. We assume the unknown Bernoulli parameters are independent observations from a common distribution F, and the objective is to provide strategies for selecting arms at each decision epoch so that the expected long run failure rate is minimized. In the first part of this study, we assume the common distribution F is arbitrary but known. We introduce three strategies proposed by Berry et al. (1996) and show that they asymptotically minimize the expected long run failure rate. Numerical results from computer simulations are also provided to evaluate the performance of the three strategies. In the second part of this study, we assume the common distribution F is unknown. For this setting, we propose a strategy called the “empirical non-recalling m-run strategy” and prove that this strategy is asymptotically optimal. Numerical results from computer simulations will also be provided to evaluate the proposed strategy and two other strategies by Herschkorn et al. (1995). |
| Relation: | 基礎研究
學術補助
研究期間:9908~ 10007
研究經費:429仟元 |
| Data Type: | report |
| Appears in Collections: | [統計學系] 國科會研究計畫
|
Files in This Item:
| File |
Size | Format | |
|---|
| 99211M004.pdf | 3387Kb | Adobe PDF2 | 1237 | View/Open |
|
All items in 政大典藏 are protected by copyright, with all rights reserved.
|
著作權政策宣告 Copyright Announcement1.本網站之數位內容為國立政治大學所收錄之機構典藏，無償提供學術研究與公眾教育等公益性使用，惟仍請適度，合理使用本網站之內容，以尊重著作權人之權益。商業上之利用，則請先取得著作權人之授權。
The digital content of this website is part of National Chengchi University Institutional Repository. It provides free access to academic research and public education for non-commercial use. Please utilize it in a proper and reasonable manner and respect the rights of copyright owners. For commercial use, please obtain authorization from the copyright owner in advance.
2.本網站之製作，已盡力防止侵害著作權人之權益，如仍發現本網站之數位內容有侵害著作權人權益情事者，請權利人通知本網站維護人員(
nccur@nccu.edu.tw)，維護人員將立即採取移除該數位著作等補救措施。
NCCU Institutional Repository is made to protect the interests of copyright owners. If you believe that any material on the website infringes copyright, please contact our staff(
nccur@nccu.edu.tw). We will remove the work from the repository and investigate your claim.
