1 \documentclass{article}
2 %\usepackage{prentcsmacro}
4 \usepackage[a4paper]{geometry}
5 \geometry{hmargin=3cm, vmargin=3cm }
7 \usepackage[latin1]{inputenc}
8 \usepackage[T1]{fontenc}
9 \usepackage[english]{babel}
10 \usepackage{amsmath,amssymb,latexsym,eufrak,euscript}
11 \usepackage{subfigure,pstricks,pst-node,pst-coil}
15 \usepackage{pgflibrarysnakes}
19 \usetikzlibrary{arrows}
20 \usetikzlibrary{automata}
21 \usetikzlibrary{snakes}
22 \usetikzlibrary{shapes}
24 %% \setlength{\oddsidemargin}{15mm}
25 %% \setlength{\evensidemargin}{15mm} \setlength{\textwidth}{140mm}
26 %% \setlength{\textheight}{219mm} \setlength{\topmargin}{5mm}
27 \newtheorem{theorem}{Theorem}
28 %\newtheorem{definition}[theorem]{Definition}
29 % %\newtheorem{defis}[thm]{D\'efinitions}
30 \newtheorem{example}[theorem]{Example}
31 % %\newtheorem{Exes}[thm]{Exemples}
32 \newtheorem{lemma}[theorem]{Lemma}
33 \newtheorem{proposition}[theorem]{Proposition}
34 \newtheorem{construction}[theorem]{Construction}
35 \newtheorem{corollary}[theorem]{Corollary}
36 % \newtheorem{algor}[thm]{Algorithm}
37 %\newtheorem{propdef}[thm]{Proposition-D\'efinition}
38 \newcommand{\mlabel}[1]{\label{#1}\marginpar{\fbox{#1}}}
39 \newcommand{\flsup}[1]{\stackrel{#1}{\longrightarrow}}
41 \newcommand{\stirlingtwo}[2]{\genfrac{\lbrace}{\rbrace}{0pt}{}{#1}{#2}}
42 \newcommand{\stirlingone}[2]{\genfrac{\lbrack}{\rbrack}{0pt}{}{#1}{#2}}
47 \newline \vspace{-0.1em}
48 \begin{quote}\begin{rm}}
49 {\end{rm}\end{quote}\end{algor}\vspace{-1.5em}\vspace{2em}}
50 %\null \hfill $\diamondsuit$ \par\medskip \vspace{1em}}
56 %\null \hfill $\triangledown$ \par\medskip}
57 %\null \hfill $\triangledown$ \par\medskip \vspace{1em}}
61 \newenvironment{proof}
62 { \noindent {\sc Proof.\/} }
63 {\null \hfill $\Box$ \par\medskip \vspace{1em}}
67 \newcommand {\tv}[1] {\lVert #1 \rVert_{\rm TV}}
72 \def \ts {\tau_{\rm stop}}
76 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
77 \section{Mathematical Backgroung}
81 Let $\pi$, $\mu$ be two distribution on a same set $\Omega$. The total
82 variation distance between $\pi$ and $\mu$ is denoted $\tv{\pi-\mu}$ and is
84 $$\tv{\pi-\mu}=\max_{A\subset \Omega} |\pi(A)-\mu(A)|.$$ It is known that
85 $$\tv{\pi-\mu}=\frac{1}{2}\sum_{x\in\Omega}|\pi(x)-\mu(x)|.$$ Moreover, if
86 $\nu$ is a distribution on $\Omega$, one has
87 $$\tv{\pi-\mu}\leq \tv{\pi-\nu}+\tv{\nu-\mu}$$
89 Let $P$ be the matrix of a markov chain on $\Omega$. $P(x,\cdot)$ is the
90 distribution induced by the $x$-th row of $P$. If the markov chain induced by
91 $P$ has a stationary distribution $\pi$, then we define
92 $$d(t)=\max_{x\in\Omega}\tv{P^t(x,\cdot)-\pi},$$
95 $$t_{\rm mix}(\varepsilon)=\min\{t \mid d(t)\leq \varepsilon\}.$$
98 $$t_{\rm mix}(\varepsilon)\leq \lceil\log_2(\varepsilon^{-1})\rceil t_{\rm mix}(\frac{1}{4})$$
102 %% A {\it coupling} with transition matrix $P$ is a process $(X_t,Y_t)_{t\geq 0}$
103 %% such that both $(X_t)$ and $(Y_t)$ are markov chains of matric $P$; moreover
104 %% it is required that if $X_s=Y_s$, then for any $t\geq s$, $X_t=Y_t$.
105 %% A results provides that if $(X_t,Y_t)_{t\geq 0}$ is a coupling, then
106 %% $$d(t)\leq \max_{x,y} P_{x,y}(\{\tau_{\rm couple} \geq t\}),$$
107 %% with $\tau_{\rm couple}=\min_t\{X_t=Y_t\}$.
110 Let $(X_t)_{t\in \mathbb{N}}$ be a sequence of $\Omega$ valued random
111 variables. A $\mathbb{N}$-valued random variable $\tau$ is a {\it stopping
112 time} for the sequence $(X_i)$ if for each $t$ there exists $B_t\subseteq
113 \omega^{t+1}$ such that $\{tau=t\}=\{(X_0,X_1,\ldots,X_t)\in B_t\}$.
115 Let $(X_t)_{t\in \mathbb{N}}$ be a markov chain and $f(X_{t-1},Z_t)$ a
116 random mapping representation of the markov chain. A {\it randomized
117 stopping time} for the markov chain is a stopping time for
118 $(Z_t)_{t\in\mathbb{N}}$. It he markov chain is irreductible and has $\pi$
119 as stationary distribution, then a {\it stationay time} $\tau$ is a
120 randomized stopping time (possibily depending on the starting position $x$),
121 such that the distribution of $X_\tau$ is $\pi$:
122 $$\P_x(X_\tau=y)=\pi(y).$$
125 If $\tau$ is a strong stationary time, then $d(t)\leq \max_{x\in\Omega}
128 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
129 \section{Random walk on the modified Hypercube}
132 Let $\Omega=\{0,1\}^N$ be the set of words of length $N$. Let $E=\{(x,y)\mid
133 x\in \Omega, y\in \Omega,\ x=y \text{ or } x\oplus y \in 0^*10^*\}$. Let $h$
134 be a function from $\Omega$ into $\{1,\ldots,N\}$.
136 We denote by $E_h$ the set $E\setminus\{(x,y)\mid x\oplus y =
137 0^{N-h(x)}10^{h(x)-1}\}$. We define the matrix $P_h$ has follows:
140 P_h(x,y)=0 & \text{ if } (x,y)\notin E_h\\
141 P_h(x,x)=\frac{1}{2}+\frac{1}{2N} & \\
142 P_h(x,x)=\frac{1}{2N} & \text{otherwise}\\
148 We denote by $\ov{h}$ the function from $\Omega$ into $\omega$ defined
149 by $x\oplus\ov{h}(x)=0^{N-h(x)}10^{h(x)-1}.$
150 The function $\ov{h}$ is said {\it square-free} if for every $x\in E$,
151 $\ov{h}(\ov{h}(x))\neq x$.
153 \begin{lemma}\label{lm:h}
154 If $\ov{h}$ is bijective and square-free, then $h(\ov{h}^{-1}(x))\neq h(x)$.
161 Let $Z$ be a random variable over
162 $\{1,\ldots,N\}\times\{0,1\}$ uniformaly distributed. For $X\in \Omega$, we
163 define, with $Z=(i,x)$,
167 f(X,Z)=X\oplus (0^{N-i}10^{i-1}) & \text{if } x=1 \text{ and } i\neq h(X),\\
168 f(X,Z)=X& \text{otherwise.}
174 %%%%%%%%%%%%%%%%%%%%%%%%%%%ù
175 \section{Stopping time}
177 An integer $\ell\in\{1,\ldots,N\}$ is said {\it fair} at time $t$ if there
178 exists $0\leq j <t$ such that $Z_j=(\ell,\cdot)$ and $h(X_j)\neq \ell$.
181 Let $\ts$ be the first time all the elements of $\{1,\ldots,N\}$
182 are fair. The integer $\ts$ is a randomized stopping time for
183 the markov chain $(X_t)$.
187 The integer $\ts$ is a strong stationnary time.
191 Let $\tau_\ell$ be the first time that $\ell$ is fair. The random variable
192 $Z_{\tau_\ell-1}$ is of the form $(\ell,\delta)$ with $\delta\in\{0,1\}$ and
193 $\delta=1$ with probability $\frac{1}{2}$ and $\delta=0$ with probability
194 $\frac{1}{2}$. Since $h(X_{\tau_\ell-1})\neq\ell$ the value of the $\ell$-th
195 bit of $X_{\tau_\ell}$ is $\delta$. Moving next in the chain, at each step,
196 the $l$-th bit is switch from $0$ to $1$ or from $1$ to $0$ each time with
197 the same probability. Therefore, for $t\geq \tau_\ell$, the
198 $\ell$-th bit of $X_t$ is $0$ or $1$ with the same probability, proving the
202 \begin{proposition} \label{prop:stop}
203 If $\ov{h}$ is bijective and square-free, then
204 $E[\ts]\leq 8N^2+ N\ln (N+1)$.
207 For each $x\in \Omega$ and $\ell\in\{1,\ldots,N\}$, let $S_{x,\ell}$ be the
208 random variable counting the number of steps done until reaching from $x$ a state where
209 $\ell$ is fair. More formaly
210 $$S_{x,\ell}=\min \{m \geq 1\mid h(X_m)\neq \ell\text{ and }Z_m=\ell\text{ and } X_0=x\}.$$
213 $$\lambda_h=\max_{x,\ell} S_{x,\ell}.$$
216 \begin{lemma}\label{prop:lambda}
217 If $\ov{h}$ is a square-free bijective function, then one has $E[\lambda_h]\leq 8N^2.$
221 For every $X$, every $\ell$, one has $\P(S_{X,\ell}\leq 2)\geq
222 \frac{1}{4N^2}$. Let $X_0=X$. Indeed, if $h(X)\neq \ell$, then
223 $\P(S_{X,\ell}=1)=\frac{1}{2N}\geq \frac{1}{4N^2}$. If $h(X)=\ell$, then
224 $\P(S_{X,\ell}=1)=0$. But in this case, intutively, it is possible to move
225 from $X$ to $\ov{h}^{-1}(X)$ (with probability $\frac{1}{2N}$). And in
226 $\ov{h}^{-1}(X)$ the $l$-th bit is switchable. More fromaly,
227 since $\ov{h}$ is square-free,
228 $\ov{h}(x)=\ov{h}(\ov{h}(\ov{h}^{-1}(X)))\neq \ov{h}^{-1}(X)$. It follows
229 that $(X,\ov{h}^{-1}(X))\in E_h$. We thus have
230 $P(X_1=\ov{h}^{-1}(X))=\frac{1}{2N}$. Now, by Lemma~\ref{lm:h},
231 $h(\ov{h}^{-1}(X))\neq h(X)$. Therefore $\P(S_{x,\ell}=2\mid
232 X_1=\ov{h}^{-1}(X))=\frac{1}{2N}$, proving that $\P(S_{x,\ell}\leq 2)\geq
235 Therefore, $\P(S_{x,\ell}\geq 3)\leq 1-\frac{1}{4N^2}$. By induction, one
236 has, for every $i$, $\P(S_{x,\ell}\geq 2i+1)\leq
237 \left(1-\frac{1}{4N^2}\right)^i$.
239 since $S_{X,\ell}$ is positive, it is known~\cite[lemma 2.9]{proba}, that
240 $$E[S_{X,\ell}]=\sum_{i=1}^{+\infty}\P(S_{X,\ell}\geq i).$$
241 Since $\P(S_{X,\ell}\geq i)\geq \P(S_{X,\ell}\geq i+1)$, one has
242 $$E[S_{X,\ell}]=\sum_{i=1}^{+\infty}\P(S_{X,\ell}\geq i)\leq
243 \P(S_{X,\ell}\geq 1)+\P(S_{X,\ell}\geq 2)+2 \sum_{i=1}^{+\infty}\P(S_{x,\ell}\geq 2i).$$
245 $$E[S_{x,\ell}]\leq 1+1+2
246 \sum_{i=1}^{+\infty}\left(1-\frac{1}{4N^2}\right)^i=2+2(4N^2-1)=8N^2,$$
247 which concludes the proof.
250 Let $\ts^\prime$ be the first time that there are exactly $N-1$ fair
253 \begin{lemma}\label{lm:stopprime}
254 One has $E[\ts^\prime]\leq N \ln (N+1).$
258 This is a classical Coupon Collector's like problem. Let $W_i$ be the
259 random variable counting the number of moves done in the markov chain while
260 we had exactly $i-1$ fair bits. One has $\ts^\prime=\sum_{i=1}^{N-1}W_i$.
261 But when we are at position $x$ with $i-1$ fair bits, the probability of
262 obtaining a new fair bit is either $1-\frac{i-1}{N}$ if $h(x)$ is fair,
263 or $1-\frac{i-2}{N}$ if $h(x)$ is not fair. It follows that
264 $E[W_i]\leq \frac{N}{N-i+2}$. Therefore
265 $$E[\ts^\prime]=\sum_{i=1}^{N-1}E[W_i]\leq N\sum_{i=1}^{N-1}
266 \frac{1}{N-i+2}=N\sum_{i=3}^{N+1}\frac{1}{i}.$$
268 But $\sum_{i=1}^{N+1}\frac{1}{i}\leq 1+\ln(N+1)$. It follows that
269 $1+\frac{1}{2}+\sum_{i=3}^{N+1}\frac{1}{i}\leq 1+\ln(N+1).$
271 $E[\ts^\prime]\leq N (-\frac{1}{2}+\ln(N+1))\leq N\ln(N+1)$.
274 One can now prove Proposition~\ref{prop:stop}.
277 One has $\ts\leq \ts^\prime+\lambda_h$. Therefore,
278 Proposition~\ref{prop:stop} is a direct application of
279 lemma~\ref{prop:lambda} and~\ref{lm:stopprime}.