数学物理学报 ›› 2014, Vol. 34 ›› Issue (6): 1554-1577.

• 论文 • 上一篇    下一篇

一类偏向删点及顶点有限制的随机图上的相变

王彬   

  1. 桂林理工大学理学院 广西桂林 |541004
  • 收稿日期:2013-03-27 修回日期:2014-06-18 出版日期:2014-12-25 发布日期:2014-12-25
  • 基金资助:

    国家自然科学基金(11001137, 11401127)、广西自然科学基金(2014GXNSFCA118015)和桂林理工大学科研启动金资助

Phase Transition in A Random Graph Process with Preferential Deletion and Limiting on Vertices

 WANG Bin   

  1. College of Science, Guilin University of Technology, Guilin Guangxi |541004
  • Received:2013-03-27 Revised:2014-06-18 Online:2014-12-25 Published:2014-12-25
  • Supported by:

    国家自然科学基金(11001137, 11401127)、广西自然科学基金(2014GXNSFCA118015)和桂林理工大学科研启动金资助

摘要:

固定α0∈[0,1)及β∈[0, 1/2). 该文引入如下随机图过程(Gt)t≥1: 设在时刻1及2已存在图G1=G2,  其中G1的顶点为v1, v2 且它们之间有2条边相连. 当t≥3时, Gt 定义如下: (i) Gt-1中任意顶点v不活跃的概率为α0. 顶点不活跃意味着其不能与t 时刻新增加的顶点相连. 此概率独立于自己以及其他顶点t-1之前的状态;  (ii) 以概率1-β增加一个新顶点vt. 在Gt-1中以概率dw(t-1)/∑vdv(t-1) 选一顶点w, 其中dw(t-1)表wGt-1中的度. 若w是活跃的则在vtw之间连1条边, 否则在vt上加个环;  (iii) 以概率βGt-1中删去一顶点u, 其中u被选中的概率为(1-du(t-1)/∑vdv(t-1))/(nt-1-1).  此处, nt-1Gt-1的顶点个数. 令Nk(t)表Gt中度为k的顶点个数. 该文证明了Gt 度分布的期望在2β/ 1-α0=1附近存在一相变:  当2β/1-α0 >1时, Nk(t)/t 的期望是呈指数衰减的;  2β/1-α0<1时, Nk(t)/t 的期望是呈幂律衰减的.

关键词: 度序列, 偏向删点, 随机图过程, 活跃

Abstract:

The following random graph process Gt is introduced. Assuming that at the time-step 1 and 2 there has been a graph G1=G2 consisting of vertices v1, v2 and 2 edges between them. At time-step t≥3, Gt is defined as follows: (i) each vertex vGt-1 is inactive with probability α0,  independently its and other vertices' statuses before t-1. The inactiveness of  means it can not being connected by more edges; (ii)  with probability 1-β a new vertex vt is added along with one edge connected to it. Then a vertex  wi is chosen with probability proportional to its degree. If wi is active then connect vt to wi. Otherwise, the edge of vt is connected to itself; (iii) with probability β a vertex is deleted preferentially from the network.  It is proved that there is a phase transition on expected degree distribution when 2β//(1-α0) is in the vicinity of 1. In the supercritical regime (2β/(1-α0)>1), its expected fraction of vertices with degree k decays exponentially.  While in the subcritical regime (2β/(1-α0)<1), the expected fraction of vertices with degree k decreases asymptotically as a power law.

Key words: Degree sequence, Preferential deletion,  Random graph process, Inactive

中图分类号: 

  • 05C80