Let (X, p ) be a probability space equipped with a kernel K : X x X → R +. For example, K is the transition matrix of a finite Markov chain or the adjacency matrix of a finite connected graph.; In this thesis we use functional-analytic methods to study discrete isoperimetric inequalities of the form p+A≥ cIpA ,A⊂X,c0, connecting suitably defined "surface" measures p +(A) and "volumes" p (A) via some non-negative functions I on [0,1]. The constant c is usually referred to as an isoperimetric constant. When estimating the isoperimetric constants, various functional constants appear naturally. These include the spectral gap, the log-Sobolev constant and others, not defined previously. Special attention is given to Cartesian products of Markov chains and graphs.
展开▼
机译:令(X,p)是配备有内核K的概率空间:X x X→R +。例如,K是有限马尔可夫链的转移矩阵或有限连通图的邻接矩阵。在本文中,我们使用泛函分析方法研究了形式为p +A≥cIpA,A⊂X,c> 0的离散等距不等式,并连接了适当定义的“表面”度量p +(A)和“体积” p(A )通过[0,1]上的一些非负函数I。常数c通常称为等压常数。估计等电常数时,自然会出现各种功能常数。这些包括光谱间隙,log-Sobolev常数和其他未定义的值。特别注意马尔可夫链和图的笛卡尔积。
展开▼