文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Proposed Running Head
侵权投诉

Proposed Running Head

itai and shachnai, adaptive source routing. Address for Proofs:

Itai, Shachnai: Adaptive Source Routing..

25

Appendix

Proof of Lemma 1: The network topology implies, that any link emanating from s belongs to a

single s-t path. The proof is by interchange argument: Since any path converging network is an sp graph, we use Theorem 1 to identify an optimal strategy with a unique order of trials on the network links. Assume that M (G)= (m1;:::; mn) is a complete order on paths for an optimal strategy S on G, then M (G) contains every s-t path in G. Let mi; mi+1 be two consecutive elements in M (G), with the success probabilities pm; pm+1 respectively, then it is su cient to show, that pm< pm+1=) S is not optimal: Denote by ProbS (T> r) the probability, that for a given network, the number of trials of paths under the strategy S is higher than r, then i i i i

ES (G; p)=

n?1 X r=0

ProbS (T> r):

(12) (13)

We denote by S 0 the strategy, which switches mi and mi+1, i.e. M

= (m1;:::; mi+1; mi;:::; mn):

We handle the following two cases separately: 1. mi and mi+1 are disjoint, then it follows from 15], that

ES (G; p)? ES 0 (G; p)> 0: 2. If mi and mi+1 are dependent, then let

(14) l)

mi= (ei;:::; eik ) 1

mi= (e i +1

(+1)1

;:::; e i

(+1)

:

For some 1 j min(l; j ), the last j links in mi; mi+1 are identical. We use the notation pir; p(i+1)r for the success probabilities on eir 1 r k, e(i+1)r 1 r j respectively, then pi(k?j)= p(i+1)(l?j);:::; pik= p(i+1)l; and since failure probabilities on the links are independent,

pm+1> pm=) i i

l?j?1 Y r=1

pi

(+1)

r>

k?j?1 Y r=1

pir;

thus, conditioning on the location of the rst fault on each of the paths under the strategies S and S 0, we nd ES (G; p)? ES 0 (G; p)> 0:

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email