BZPEER

preprint2016arXiv

A computable bound of the essential spectral radius of finite range Metropolis--Hastings kernels

Let $π$ be a positive continuous target density on $\mathbb{R}$ . Let $P$ be the Metropolis-Hastings operator on the Lebesgue space $\mathbb{L}^2(π)$ corresponding to a proposal Markov kernel $Q$ on $\mathbb{R}$ . When using the quasi-compactness method to estimate the spectral gap of $P$ , a mandatory first step is to obtain an accurate bound of the essential spectral radius $r\_{ess}(P)$ of $P$ . In this paper a computable bound of $r\_{ess}(P)$ is obtained under the following assumption on the proposal kernel: $Q$ has a bounded continuous density $q(x,y)$ on $\mathbb{R}^2$ satisfying the following finite range assumption : $|u| \textgreater{} s \, \Rightarrow\, q(x,x+u) = 0$ (for some $s\textgreater{}0$). This result is illustrated with Random Walk Metropolis-Hastings kernels.

preprint2016arXiv

Additional material on bounds of $\ell^2$ -spectral gap for discrete Markov chains with band transition matrices

We analyse the $\ell^2(π)$ -convergence rate of irreducible and aperiodic Markov chains with $N$ -band transition probability matrix $P$ and with invariant distribution $π$ . This analysis is heavily based on: first the study of the essential spectral radius $r\_{ess}(P\_{|\ell^2(π)})$ of $P\_{|\ell^2(π)}$ derived from Hennion's quasi-compactness criteria; second the connection between the spectral gap property (SG $\_2$ ) of $P$ on $\ell^2(π)$ and the $V$ -geometric ergodicity of $P$ . Specifically, (SG $\_2$ ) is shown to hold under the condition \[α\_0 := \sum\_{{m}=-N}^N \limsup\_{i\rightarrow +\infty} \sqrt{P(i,i+{m})\, P^*(i+{m},i)}\ \textless{}\, 1. \] Moreover $r\_{ess}(P\_{|\ell^2(π)}) \leq α\_0$ . Simple conditions on asymptotic properties of $P$ and of its invariant probability distribution $π$ to ensure that $α\_0\textless{}1$ are given. In particular this allows us to obtain estimates of the $\ell^2(π)$ -geometric convergence rate of random walks with bounded increments. The specific case of reversible $P$ is also addressed. Numerical bounds on the convergence rate can be provided via a truncation procedure. This is illustrated on the Metropolis-Hastings algorithm.

preprint2015arXiv

Computable bounds of ${\ell}^2$ -spectral gap for discrete Markov chains with band transition matrices

We analyse the $\ell^2(π)$ -convergence rate of irreducible and aperiodic Markov chains with $N$ -band transition probability matrix $P$ and with invariant distribution $π$ . This analysis is heavily based on: first the study of the essential spectral radius $r\_{ess}(P\_{|\ell^2(π)})$ of $P\_{|\ell^2(π)}$ derived from Hennion's quasi-compactness criteria; second the connection between the Spectral Gap property (SG $\_2$ ) of $P$ on $\ell^2(π)$ and the $V$ -geometric ergodicity of $P$ . Specifically, (SG $\_2$ ) is shown to hold under the condition $ α\_0 := \sum\_{{m}=-N}^N \limsup\_{i\rightarrow +\infty} \sqrt{P(i,i+{m})\, P^*(i+{m},i)}\ \textless{}\, 1 $ Moreover $r\_{ess}(P\_{|\ell^2(π)}) \leq α\_0$ . Effective bounds on the convergence rate can be provided from a truncation procedure.

preprint2014arXiv

Approximating Markov chains and V-geometric ergodicity via weak perturbation theory

Let $P$ be a Markov kernel on a measurable space $\X$ and let $V:\X\r[1,+\infty)$. This paper provides explicit connections between the $V$ -geometric ergodicity of $P$ and that of finite-rank nonnegative sub-Markov kernels $\Pc_k$ approximating $P$ . A special attention is paid to obtain an efficient way to specify the convergence rate for $P$ from that of $\Pc_k$ and conversely. Furthermore, explicit bounds are obtained for the total variation distance between the $P$ -invariant probability measure and the $\Pc_k$-invariant positive measure. The proofs are based on the Keller-Liverani perturbation theorem which requires an accurate control of the essential spectral radius of $P$ on usual weighted supremum spaces. Such computable bounds are derived in terms of standard drift conditions. Our spectral procedure to estimate both the convergence rate and the invariant probability measure of $P$ is applied to truncation of discrete Markov kernels on $\X:=\N$.

preprint2013arXiv

A local limit theorem for densities of the additive component of a finite Markov Additive Process

In this paper, we are concerned with centered Markov Additive Processes $\{(X_t,Y_t)\}_{t\in\T}$ where the driving Markov process $\{X_t\}_{t\in\T}$ has a finite state space. Under suitable conditions, we provide a local limit theorem for the density of the absolutely continuous part of the probability distribution of $t^{-1/2}Y_t$ given $X_0$ . The rate of convergence and the moment condition are the expected ones with respect to the i.i.d case. An application to the joint distribution of local times of a finite jump process is sketched.

preprint2013arXiv

Additional material on local limit theorem for finite Additive Markov Processes [HL13]

This paper proposes additional material to the main statements of [HL13] which are are recalled in Section~2. In particular an application of [Theorem~2.2,HL13] to Renewal Markov Processes is provided in Section~4 and a detailed checking of the assumptions of [Theorem~2.2,HL13] for the joint distribution of local times of a finite jump process is reported in Section~5. A uniform version of [Theorem~2.2,HL13] with respect to a compact set of transition matrices is given in Section~6 (see [Remark 2.4,HL13]). The basic material on the semigroup of Fourier matrices and the spectral approach used in [HL13] is recalled in Section~3 in order to obtain a good understanding of the properties involved in this uniform version.

preprint2013arXiv

Spectral analysis of Markov kernels and application to the convergence rate of discrete random walks

Let $\{X_n\}_{n\in\N}$ be a Markov chain on a measurable space $\X$ with transition kernel $P$ and let $V:\X\r[1,+\infty)$. The Markov kernel $P$ is here considered as a linear bounded operator on the weighted-supremum space $\cB_V$ associated with $V$ . Then the combination of quasi-compactness arguments with precise analysis of eigen-elements of $P$ allows us to estimate the geometric rate of convergence $ρ_V(P)$ of $\{X_n\}_{n\in\N}$ to its invariant probability measure in operator norm on $\cB_V$. A general procedure to compute $ρ_V(P)$ for discrete Markov random walks with identically distributed bounded increments is specified.

preprint2012arXiv

A uniform Berry--Esseen theorem on $M$ -estimators for geometrically ergodic Markov chains

Let $\{X_n\}_{n\ge0}$ be a $V$ -geometrically ergodic Markov chain. Given some real-valued functional $F$ , define $M_n(α):=n^{-1}\sum_{k=1}^nF(α,X_{k-1},X_k)$ , $α\in\mathcal{A}\subset \mathbb {R}$ . Consider an $M$ estimator $\hatα_n$ , that is, a measurable function of the observations satisfying $M_n(\hatα_n)\leq \min_{α\in\mathcal{A}}M_n(α)+c_n$ with $\{c_n\}_{n\geq1}$ some sequence of real numbers going to zero. Under some standard regularity and moment assumptions, close to those of the i.i.d. case, the estimator $\hatα_n$ satisfies a Berry--Esseen theorem uniformly with respect to the underlying probability distribution of the Markov chain.

preprint2012arXiv

Limit theorems for stationary Markov processes with L2-spectral gap

Let $(X_t, Y_t)_{t\in T}$ be a discrete or continuous-time Markov process with state space $X \times R^d$ where $X$ is an arbitrary measurable set. Its transition semigroup is assumed to be additive with respect to the second component, i.e. $(X_t, Y_t)_{t\in T}$ is assumed to be a Markov additive process. In particular, this implies that the first component $(X_t)_{t\in T}$ is also a Markov process. Markov random walks or additive functionals of a Markov process are special instances of Markov additive processes. In this paper, the process $(Y_t)_{t\in T}$ is shown to satisfy the following classical limit theorems: (a) the central limit theorem, (b) the local limit theorem, (c) the one-dimensional Berry-Esseen theorem, (d) the one-dimensional first-order Edgeworth expansion, provided that we have sup{t\in(0,1]\cap T : E{π,0}[|Y_t| ^α] < 1 with the expected order with respect to the independent case (up to some $\varepsilon > 0$ for (c) and (d)). For the statements (b) and (d), a Markov nonlattice condition is also assumed as in the independent case. All the results are derived under the assumption that the Markov process $(X_t)_{t\in T}$ has an invariant probability distribution $π$ , is stationary and has the $L^2(π)$ -spectral gap property (that is, $(X_t)t\in N}$ is $ρ$ -mixing in the discrete-time case). The case where $(X_t)_{t\in T}$ is non-stationary is briefly discussed. As an application, we derive a Berry-Esseen bound for the M-estimators associated with $ρ$ -mixing Markov chains.

preprint2012arXiv

Regular perturbation of V -geometrically ergodic Markov chains

In this paper, new conditions for the stability of V-geometrically ergodic Markov chains are introduced. The results are based on an extension of the standard perturbation theory formulated by Keller and Liverani. The continuity and higher regularity properties are investigated. As an illustration, an asymptotic expansion of the invariant probability measure for an autoregressive model with i.i.d. noises (with a non-standard probability density function) is obtained.

James Ledoux

Research overview

Is this your work?

10 publications

A computable bound of the essential spectral radius of finite range Metropolis--Hastings kernels

Additional material on bounds of $\ell^2$ -spectral gap for discrete Markov chains with band transition matrices

Computable bounds of ${\ell}^2$ -spectral gap for discrete Markov chains with band transition matrices

Approximating Markov chains and V-geometric ergodicity via weak perturbation theory

A local limit theorem for densities of the additive component of a finite Markov Additive Process

Additional material on local limit theorem for finite Additive Markov Processes [HL13]

Spectral analysis of Markov kernels and application to the convergence rate of discrete random walks

A uniform Berry--Esseen theorem on $M$ -estimators for geometrically ergodic Markov chains

Limit theorems for stationary Markov processes with L2-spectral gap

Regular perturbation of V -geometrically ergodic Markov chains

See the researcher in context

Building this map preview

James Ledoux

Research overview

Is this your work?

10 publications

A computable bound of the essential spectral radius of finite range Metropolis--Hastings kernels

Additional material on bounds of ℓ2\ell^2ℓ2-spectral gap for discrete Markov chains with band transition matrices

Computable bounds of ℓ2{\ell}^2ℓ2-spectral gap for discrete Markov chains with band transition matrices

Approximating Markov chains and V-geometric ergodicity via weak perturbation theory

A local limit theorem for densities of the additive component of a finite Markov Additive Process

Additional material on local limit theorem for finite Additive Markov Processes [HL13]

Spectral analysis of Markov kernels and application to the convergence rate of discrete random walks

A uniform Berry--Esseen theorem on MMM-estimators for geometrically ergodic Markov chains

Limit theorems for stationary Markov processes with L2-spectral gap

Regular perturbation of V -geometrically ergodic Markov chains

See the researcher in context

Building this map preview

Additional material on bounds of $\ell^2$ -spectral gap for discrete Markov chains with band transition matrices

Computable bounds of ${\ell}^2$ -spectral gap for discrete Markov chains with band transition matrices

A uniform Berry--Esseen theorem on $M$ -estimators for geometrically ergodic Markov chains