\documentclass{article} \usepackage{amsmath,amssymb} \setlength{\oddsidemargin}{0.25 in} \setlength{\evensidemargin}{-0.25 in} \setlength{\topmargin}{-0.6 in} \setlength{\textwidth}{6.5 in} \setlength{\textheight}{8.5 in} \setlength{\headsep}{0.75 in} \setlength{\parindent}{0 in} \setlength{\parskip}{0.1 in} \newcounter{lecnum} % % The following macro is used to generate the header. % \newcommand{\lecture}[4]{ \pagestyle{myheadings} \thispagestyle{plain} \newpage \setcounter{lecnum}{#1} \setcounter{page}{1} \noindent \begin{center} \framebox{ \vbox{\vspace{2mm} \hbox to 6.28in { {{\bf 18.317~Combinatorics, Probability, and Computations on Groups} \hfill #2} } \vspace{4mm} \hbox to 6.28in { {\Large \hfill Lecture #1 \hfill} } \vspace{2mm} \hbox to 6.28in { {\it Lecturer: #3 \hfill Scribe: #4} } \vspace{2mm}}} \end{center} \markboth{Lecture #1: #2}{Lecture #1: #2} \vspace*{4mm} } \input{epsf} %usage: \fig{LABEL}{FIGURE-HEIGHT}{CAPTION}{FILENAME} \newcommand{\fig}[4]{ \begin{figure} \setlength{\epsfysize}{#2} \centerline{\epsfbox{#4}} \caption{#3} \label{#1} \end{figure} } \newtheorem{theorem}{Theorem} \newtheorem{lemma}[theorem]{Lemma} \newtheorem{proposition}[theorem]{Proposition} \newtheorem{claim}[theorem]{Claim} \newtheorem{corollary}[theorem]{Corollary} \newtheorem{definition}[theorem]{Definition} \newenvironment{proof}{{\bf Proof:}}{\hfill\rule{2mm}{2mm}} \newenvironment{mproof}[1]{{\bf #1}}{\hfill\rule{2mm}{2mm}} \def\E#1{{\rm E}[#1]} \def\i{\hspace*{5mm}} \newcounter{lineno} \setcounter{lineno}{0} \def\n{\addtocounter{lineno}{1} \hbox to 7mm{\hfill \arabic{lineno}:}\hspace{2mm}} \def\resetn{\setcounter{lineno}{0}} \def\tqbfk{\tqbf_k^\exists} \def\ep{\epsilon} \def\erd{Erd\H{o}s} \begin{document} \lecture{21}{November 2, 2001}{Igor Pak}{B. Virag} \newcommand{\ee}{\mathcal E} \newcommand{\ev}{\mbox{\bf E}} \newcommand{\pr}{\mbox{\bf P}} \section{Dirichlet forms and mixing time} Let $G$ be a finite group, and let $V$ be the vector space of real-valued functions from $G$. There is a natural inner product on this space $$ \langle \phi, \varphi \rangle = \sum_{x\in G} \varphi(x) \psi(x) = |G| \ev \varphi(X) \psi(X) $$ where $X$ is chosen from $g$ according to uniform measure. Let $\pi$ be a probability distribution on $G$, and let $P_\pi$ denote the transition kernel of the random walk $\{X_n\}$ that moves from $x$ to $xy$ in every step, where $y$ is distributed according to $\pi$. Just like any other transition kernel, $P_\pi$ acts on the space of functions on $G$ as follows $$ \left[P_\pi \varphi\right](x) =\sum_{y \in G} \varphi(xy) \pi(y) =\ev[\varphi(X_1)\ | \ X_0=x]. $$ Define the support of $\varphi\in V$ in the usual way, $$ \mbox{supp} (\varphi) = \{x\in G \ |\ \varphi(x) \not=0\}. $$ We now define the Dirichlet form $$ \ee_\pi(\varphi, \varphi)=\langle(I-P_\pi)\varphi, \varphi\rangle = |G| \ev \left[ \left(\varphi(X_0)-\varphi(X_1)\right) \varphi (X_0)\right] $$ where now $X_0$ is chosen according to uniform distribution on $G$. Note that if $Z_0,\ Z_1$ are real-valued random variables with the same distribution, then $$ \ev [(Z_0-Z_1)Z_0] = {1\over 2} \ev (Z_0-Z_1)^2. $$ Since the uniform distribution is stationary with respect to convolution, $X_0$ and $X_1$ have the same distribution, and we may apply this to $\varphi(X_0),\ \varphi(X_1)$ to get the alternative formula for the Dirichlet form $$ \ee_\pi(\varphi,\varphi) = {|G|\over 2} \ev (\varphi(X_0)-\varphi(X_1))^2 = {1 \over 2}\sum_{x,y\in G} (\varphi(x)-\varphi(xy))^2 \pi(y). $$ Now let $\Gamma=\Gamma(G,S)$ be a Cayley graph of $G$ with respect to a symmetric generator set $S$. Let $\gamma$ be a function that assigns to each $y\in G$ a path from the identity to $y$ in $\Gamma$. We will assume that $\gamma$ is geodesic, that is its values are shortest paths. Let $ \mu_s(y)=\mu_s(y,\gamma) $ denote the number of times a generator $s$ appears in the decomposition \begin{equation}\label{deco} y= s_1 s_2 \ldots s_\ell \end{equation} along the path $\gamma(y)$. For $C\subseteq G$ Let $$ N_\gamma(s,C)= \max_{x,y\in C} \mu_s(x^{-1}y). $$ We have a following version of a theorem by Diaconis and Saloff-Coste. \begin{theorem}[Comparison of Dirichlet forms] Let $C\subseteq G$, let $\overline C = C \cup \partial C$, and let $d={\rm diam}(\overline C)$. Consider $\pi,\ \tilde \pi$ symmetric probability distributions on $G$, and let $S\subseteq{\rm supp}(\pi)$. Then $$ \ee_\pi (\varphi, \varphi) \ge {1 \over A}\ee_{\tilde \pi} (\varphi, \varphi) $$ where $$ A = d \max_{s\in S} {N_\gamma (s, \overline C) \over \pi(s)}. $$ \end{theorem} \begin{proof} Let $y\in G$, and write $y$ in the form (\ref{deco}). We can write $$ \varphi(x) - \varphi (xy) = [\varphi(xs)-\varphi(xs_1)] + \ldots + [\varphi(xs_1 \ldots s_{\ell-1} - \varphi (xy)]. $$ It follows, for example, by the Cauchy-Schwarz inequality that $$ (\varphi(x) - \varphi (xy))^2 \le \ell^* \sum_{i=1}^{\ell} (\varphi(xs_1\ldots s_{i-1})- \varphi(xs_1 \ldots s_i))^2 $$ where $\ell^*$ is the number of nonzero terms in the sum, and is bounded above by $d=$diam$(\overline C)$, since $\gamma$ is geodesic. Summing this inequality over $x\in G$ we get $$ \sum_{x\in G} (\varphi(x) - \varphi (xy)) ^2 \le d \sum_{z\in G, s\in S } N_\gamma(s, \overline C)(\varphi(z)- \varphi(zs))^2. $$ Since this holds for all $y\in G$, we may average the left hand side with respect to $y$ with weights $\tilde \pi(y)$ to get $$ \sum_{x,y\in G} (\varphi(x)-\varphi(xy))^2 \tilde \pi(y) \le d \sum_{z\in G,s\in S} N_\gamma(s,\overline C)(\varphi(z)-\varphi(zs))^2 $$ Finally, the right hand side is clearly bounded by $$ A \sum_{z\in G, s\in S} (\varphi(z)-\varphi(zs))^2 \pi(s) $$ and the statement of the theorem follows. \end{proof} By the way, the symmetry assumption for $S$ was not used. \end{document}