Distributed Averaging in Opinion Dynamics

Distributed Averaging in Opinion Dynamics
复制标题

意见动态中的分布式平均

DOI:
10.1145/3583668.3594593
复制
发表时间:
2023
期刊:
--
影响因子:
--
通讯作者:
Berenbrink P
Berenbrink P
中科院分区:
--
文献类型:
--
作者:
Berenbrink P

文献摘要

参考文献

相似文献

我们考虑任意图上的两个简单的异步意见动态,其中图的每个节点都有一个初始值u(0)。在第一个过程中,我们称之为节点模型,在每个时间步长≥ 0时,一个随机节点和它的邻居的k个随机样本1,2,.,则选择了。然后,u更新其当前值<$u(t)到[EQUATION],其中α ∈(0,1)和k ≥ 1是过程的参数。在第二个过程中,称为边模型,在每一步中选择一个随机的相邻节点对(u,n),然后nodeu将其值等效地更新为节点模型,其中k = 1,n作为所选的邻居。对于两个过程,所有节点的值收敛到相同的值F,这是一个随机变量,取决于每一步中所做的随机选择。对于节点模型和规则图,以及对于边模型和任意图,F的期望是初始值的平均值[方程]。对于节点模型和非正则图,F的期望是初始值的度加权平均。我们考虑了F的集中度,并证明了正则图F的方差的紧界。我们表明,当初始值不依赖于节点的数量,那么方差是可以忽略不计的,因此节点能够估计的初始平均值的节点值。有趣的是,这种变化并不依赖于图形结构。为了证明,我们引入了我们的过程和两个相关的随机游动过程之间的对偶。我们还分析了两个模型和任意图的收敛时间,显示了使所有节点值彼此"ε-接近"所需的时间T ε的界限。在对初值分布的一些假设下,我们的界是渐近紧的。
We consider two simple asynchronous opinion dynamics on arbitrary graphs where every nodeuof the graph has an initial valueξu(0). In the first process, which we call theNodeModel, at each time stept≥ 0, a random nodeuand a random sample ofkof its neighboursυ1,υ2, ... ,υkare selected. Then,uupdates its current valueξu(t) to [EQUATION], whereα∈ (0, 1) andk≥ 1 are parameters of the process. In the second process, called theEdgeModel, at each step a random pair of adjacent nodes (u, υ) is selected, and then nodeuupdates its value equivalently to theNodeModelwithk =1 andυas the selected neighbour.For both processes, the values of all nodes converge to the same valueF, which is a random variable depending on the random choices made in each step. For theNodeModeland regular graphs, and for theEdgeModeland arbitrary graphs, the expectation ofFis the average of the initial values [EQUATION]. For theNodeModeland non-regular graphs, the expectation ofFis the degree-weighted average of the initial values.Our results are two-fold. We consider the concentration ofFand show tight bounds on the variance ofFfor regular graphs. We show that when the initial values do not depend on the number of nodes, then the variance is negligible, and hence the nodes are able to estimate the initial average of the node values. Interestingly, this variance does not depend on the graph structure. For the proof we introduce a duality between our processes and a process of two correlated random walks. We also analyse the convergence time for both models and for arbitrary graphs, showing bounds on the timeTεrequired to make all node values 'ε-close' to each other. Our bounds are asymptotically tight under some assumptions on the distribution of the initial values.
关于平均过程的讲座
DOI: 10.1214/11-ps184
发表时间: 2012
影响因子: 1.6
作者:
D. Aldous;D. Lanoue
通讯作者: D. Lanoue
DOI: 10.1007/978-3-662-43951-7_37
发表时间: 2014-04
期刊: --
影响因子: --
作者:
C. Cooper;Robert Elsässer;T. Radzik
通讯作者: C. Cooper;Robert Elsässer;T. Radzik
DOI: --
发表时间: 2020
期刊: Sigact News
影响因子: --
作者:
L. Becchetti;A. Clementi;Emanuele Natale
通讯作者: Emanuele Natale
DOI: --
发表时间: 2015
期刊: SIAM journal on computing (Print)
影响因子: --
作者:
B. Chazelle
通讯作者: B. Chazelle
DOI: --
发表时间: 2018
期刊: Algorithmica
影响因子: 1.1
作者:
Dimitris Fotakis;Vardis Kandiros;Vasilis Kontonis;Stratis Skoulakis
通讯作者: Stratis Skoulakis