arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Author

Chii-Ruey Hwang

16 papers · 1980–2018

1980: 11981: 01982: 21983: 11984: 01985: 01986: 11987: 01988: 01989: 01990: 01991: 01992: 11993: 21994: 01995: 01996: 01997: 01998: 01999: 02000: 02001: 02002: 02003: 02004: 02005: 22006: 02007: 02008: 02009: 02010: 12011: 02012: 12013: 02014: 22015: 12016: 02017: 02018: 1
1980peak 2 in 19822018

Frequent co-authors

Papers

showing 1–16 of 16 RSS

Nonparametric Maximum Likelihood Estimation by the Method of Sieves

Stuart Geman, Chii-Ruey Hwang

Statistics and Probability · The Annals of Statistics · · cited 368× · DOI ↗ · more like this ·

Maximum likelihood estimation often fails when the parameter takes values in an infinite dimensional space. For example, the maximum likelihood method cannot be applied to the completely nonparametric estimation of a density function from an $\operatorname{iid}$ sample; the maximum of the likelihood is not attained by any density. In this example, as in many other examples, the parameter space (positive functions with area one) is too big. But the likelihood method can often be salvaged if we first maximize over a constrained subspace of the parameter space and then relax the constraint as the sample size grows. This is Grenander's "method of sieves." Application of the method sometimes leads to new estimators for familiar problems, or to a new motivation for an already well-studied technique. We will establish some general consistency results for the method, and then we will focus on…

Laplace's Method Revisited: Weak Convergence of Probability Measures

Chii-Ruey Hwang

Mathematical Physics · The Annals of Probability · · cited 158× · DOI ↗ · more like this ·

Let $Q$ be a fixed probability on the Borel $\sigma$-field in $R^n$ and $H$ be an energy function continuous in $R^n$. A set $N$ is related to $H$ by $N = \{x \mid\inf_yH(y) = H(x)\}$. Laplace's method, which is interpreted as weak convergence of probabilities, is used to introduce a probability $P$ on $N$. The general properties of $P$ are studied. When $N$ is a union of smooth compact manifolds and $H$ satisfies some smooth conditions, $P$ can be written in terms of the intrinsic measures on the highest dimensional mainfolds in $N$.

Accelerating diffusions

Chii-Ruey Hwang, Shu-Yin Hwang-Ma, Shuenn-Jyi Sheu

math.PR · · cited 141× · arXiv ↗ · more like this ·

Let U be a given function defined on R^d and π(x) be a density function proportional to \exp -U(x). The following diffusion X(t) is often used to sample from π(x), dX(t)=-\nabla U(X(t)) dt+\sqrt2 dW(t),\qquad X(0)=x_0. To accelerate the convergence, a family of diffusions with π(x) as their common equilibrium is considered, dX(t)=\bigl(-\nabla U(X(t))+C(X(t))\bigr) dt+\sqrt2 dW(t),\qquad X(0)=x_0. Let L_C be the corresponding infinitesimal generator. The spectral gap of L_C in L^2(π) (λ(C)), and the convergence exponent of X(t) to πin variational norm (ρ(C)), are used to describe the convergence rate, where λ(C)= Sup{real part of μ\dvtxμis in the spectrum of L_C, μis not zero}, {-2.8cm}ρ(C) = Inf\biggl{ρ\dvtx\int | p(t,x,y) -π(y)| dy \le g(x) e^{ρt}\biggr}.Roughly speaking, L_C is a perturbation of the self-adjoint L_0 by an antisymmetric operator C\cdot\nabla, where C is weighted…

Accelerating Gaussian Diffusions

Chii-Ruey Hwang, Shu-Yin Hwang-Ma, Shuenn‐Jyi Sheu

Statistics and Probability · The Annals of Applied Probability · · cited 109× · DOI ↗ · more like this ·

Let $\pi(x)$ be a given probability density proportional to $\exp(-U(x))$ in a high-dimensional Euclidean space $\mathbb{R}^m$. The diffusion $dX(t) = -\nabla U(X(t))dt + \sqrt 2 dW(t)$ is often used to sample from $\pi$. Instead of $-\nabla U(x)$, we consider diffusions with smooth drift $b(x)$ and having equilibrium $\pi(x)$. First we study some general properties and then concentrate on the Gaussian case, namely, $-\nabla U(x) = Dx$ with a strictly negative-definite real matrix $D$ and $b(x) = Bx$ with a stability matrix $B$; that is, the real parts of the eigenvalues of $B$ are strictly negative. Using the rate of convergence of the covariance of $X(t)$ [or together with $EX(t)$] as the criterion, we prove that, among all such $b(x)$, the drift $Dx$ is the worst choice and that improvement can be made if and only if the eigenvalues of $D$ are not identical. In fact, the convergence…

Optimal Spectral Structure of Reversible Stochastic Matrices, Monte Carlo Methods and the Simulation of Markov Random Fields

Arnoldo Frigessi, Chii-Ruey Hwang, Laurent Younès

Statistics and Probability · The Annals of Applied Probability · · cited 85× · DOI ↗ · more like this ·

In this paper we prove an optimal spectral structure theorem for stochastic matrices reversible with respect to a fixed probability measure $\pi$ on a finite set and devise a new simulation algorithm for Markov random fields. We compute the minimum value for the second largest eigenvalue of all such matrices and characterize the class of matrices for which this minimum is attained. In fact, they share a common right eigenvector that can be written in terms of $\pi$. Furthermore, by iterating this procedure, we obtain a unique matrix which is minimal with respect to the lexicographic order of the eigenvalues. We give a probabilistic interpretation of the corresponding eigenvectors. Our results allow us to devise a dynamic Monte Carlo scheme which has an optimal worst-case performance. Regarding the simulation of lattice-based Gibbs distributions, we design a modified Gibbs sampler, whose…

Convergence Rates of the Gibbs Sampler, the Metropolis Algorithm and Other Single-Site Updating Dynamics

Arnoldo Frigessi, P. Di Stefano, Chii-Ruey Hwang +1 more

Statistics and Probability · Journal of the Royal Statistical Society Series B (Statistical Methodology) · · cited 83× · DOI ↗ · more like this ·

SUMMARY Sampling from a Markov random field II can be performed efficiently via Monta Carlo methods by simulating a Markov chain that converges weakly to II. We consider a class of local updating dynamics that are reversible with respect to II. It includes the Metropolis algorithm (MA) and the Gibbs sampler (GS). We investigate the speed of weak convergence of these Markov chains in terms of their second-largest eigenvalues in absolute value. We study the general algebraic structure and then the stochastic Ising model in detail. We conclude the following: the GS is faster than locally updating twice by the MA; in the Ising case, the MA is the best at low temperature but the worst at high temperature; there are dynamics faster than the GS at high temperature. The results clear up some intuitive misconceptions.

On the Optimal Transition Matrix for Markov Chain Monte Carlo Sampling

Ting‐Li Chen, Wei-Kuo Chen, Chii-Ruey Hwang +1 more

Statistics and Probability · SIAM Journal on Control and Optimization · · cited 22× · DOI ↗ · more like this ·

Let ${\cal X}$ be a finite space and let $\pi$ be an underlying probability on ${\cal X}$. For any real-valued function $f$ defined on ${\cal X}$, we are interested in calculating the expectation of $f$ under $\pi$. Let $X_0,X_1,\ldots,X_n,\ldots$ be a Markov chain generated by some transition matrix $P$ with invariant distribution $\pi$. The time average, $\frac{1}{n}\sum_{k=0}^{n-1} f(X_k)$, is a reasonable approximation to the expectation, $E_\pi[f(X)]$. Which matrix $P$ minimizes the asymptotic variance of $\frac{1}{n}\sum_{k=0}^{n-1} f(X_k)$? The answer depends on $f$. Rather than a worst-case analysis, we will identify the set of $P$'s that minimize the average asymptotic variance, averaged with respect to a uniform distribution on $f$.

Non-parametric Decoding on Discrete Time Series and Its Applications in Bioinformatics

Fushing Hsieh, Shuchun Chen, Chii-Ruey Hwang

Artificial Intelligence · Statistics in Biosciences · · cited 7× · DOI ↗ · more like this ·

We address the question: How do we non-parametrically decode the unknown state-space vector underlying a lengthy discrete time series? The time series of concern is governed by one non-autonomous dynamics with only two internal states. This question pertinently reflects the dilemma of computing infeasibility against inferential bias found in many scientific areas. This dilemma becomes an issue when considering whether to have, or not to have likely very unrealistic structural assumptions on the state-space dynamics in most of real-world applications. To resolve this dilemma, the decoding problem is transformed into an event-intensity change-point problem without prior knowledge of the number of change-points involved. A new decoding algorithm, called Hierarchical Factor Segmentation (HFS), is proposed to achieve computability and robustness. Performance of the HFS algorithm in terms of…

ACCELERATING MONTE CARLO MARKOV PROCESSES

Chii-Ruey Hwang

Statistics and Probability · COSMOS · · cited 6× · DOI ↗ · more like this ·

Let π be a probability density proportional to exp - U(x) in S. A convergent Markov process to π(x) may be regarded as a "conceptual" algorithm. Assume that S is a finite set. Let X 0 ,X 1 ,…,X n ,… be a Markov chain with transition matrix P and invariant probability π. Under suitable condition on P, it is known that [Formula: see text] converges to π(f) and the corresponding asymptotic variance v(f, P) depends only on f and P. It is natural to consider criteria v w (P) and v a (P), defined respectively by maximizing and averaging v(f, P) over f. Two families of transition matrices are considered. There are four problems to be investigated. Some results and conjectures are given. As for the continuum case, to accelerate the convergence a family of diffusions with drift ∇U(x) + C(x) with div(C(x)exp - U(x)) = 0 is considered.

Optimal Variance Reduction for Markov Chain Monte Carlo

Lu‐Jing Huang, Yin-Ting Liao, Ting‐Li Chen +1 more

Statistics and Probability · SIAM Journal on Control and Optimization · · cited 4× · DOI ↗ · more like this ·

Markov chain Monte Carlo (MCMC) has been widely used to approximate the expectation of the statistic of a given probability measure $\pi$ on a finite set, and the asymptotic variance is a typical approach to evaluating the performance of MCMC methods. In this paper, we provide a lower bound of the worst-case analysis of the asymptotic variance over general Markov chains with invariant probability $\pi$, reversible as well as nonreversible ones, and construct an optimal transition matrix that achieves this lower bound. In fact, for any statistic $f$ to be evaluated, an MCMC with the constructed optimal transition matrix produces a smaller asymptotic variance than independent sampling.

Variance reduction for diffusions

Chii-Ruey Hwang, Raoul Normand, Sheng-Jhih Wu

math.PR · · cited 1× · arXiv ↗ · more like this ·

The most common way to sample from a probability distribution is to use Monte-Carlo methods. For distributions on a continuous state space, one can find diffusions with the target distribution as equilibrium measure, so that the state of the diffusion after a long time provides a good sample from the desired distribution. There exist many diffusions with a common equilibrium, and one would naturally like to choose those that make the convergence to equilibrium faster. One way to do this is to consider a reversible diffusion, and add to it an antisymmetric drift which preserves the invariant measure. It has been proven that, in general, the irreversible algorithm performs better than the reversible one, in that the spectral gap is larger. In the present work, asymptotic variance is used as the criterion to compare these algorithms. We first provide a general comparison result, and then…

Cramer's Theorem for Certain Ergodic Processes in Banach Spaces

Chii-Ruey Hwang, Shuenn‐Jyi Sheu

Applied Mathematics · Stochastics · · cited 1× · DOI ↗ · more like this ·

Abstract Let be i.i.d Banach space valued random variables with E exp for all Let denote the Cramer transform of . Let be an absolutely summable series of bounded linear operators with sum denoted by S. x$sub:t$esub: is defined by the moving average of . We prove that satisfies the large deviation principle with rate function i.e. I is the Cramer transform of . A specific example in R with Eexpfor t>1 only is discussed to demonstrate that the corresponding rate function is quite different from . Keywords: Banach space valued random variablesCramer's theoremergodic processlarge deviationsmoving averagespectral density

Conditioning by $\langle $equal, linear$\rangle $

Chii-Ruey Hwang

Statistics and Probability · Transactions of the American Mathematical Society · · cited 1× · DOI ↗ · more like this ·

We deal with a limit problem of regularity controlled probabilities in metric pattern theory. The probability on the generator space is given by a density function/(x, y) on which some integrability conditions are imposed. Let T denote the integral operator with kernel/. When n i.i.d. generators (Xk,Yk) are connected together to form the configuration space 6" via the regularity (EQUAL, LINEAR), i.e. "conditioning" onH, = Yk for 1 s k < n, an approximate identity is used to define the regularity controlled probability on Qn. The probabilistic effect induced by the regularity conditions on some fixed subconfiguration of a larger configuration " is described by its corresponding marginal probability within S". When n goes to infinity in a suitable way, the above mentioned marginal probability converges weakly to a limit whose density can be expressed in terms of the largest eigenvalues…

Disorder chaos in the spherical mean-field model

Wei-Kuo Chen, Hsi-Wei Hsieh, Chii-Ruey Hwang +1 more

math.PR · · arXiv ↗ · more like this ·

We consider the problem of disorder chaos in the spherical mean-field model. It is concerned about the behavior of the overlap between two independently sampled spin configurations from two Gibbs measures with the same external parameters. The prediction states that if the disorders in the Hamiltonians are slightly decoupled, then the overlap will be concentrated near a constant value. Following Guerra's replica symmetry breaking scheme, we establish this at the level of the free energy as well as the Gibbs measure irrespective the presence or absence of the external field.

On the Weak Convergence and Central Limit Theorem of Blurring and Nonblurring Processes with Application to Robust Location Estimation

Ting-Li Chen, Hironori Fujisawa, Su-Yun Huang +1 more

math.ST · · arXiv ↗ · more like this ·

This article studies the weak convergence and associated Central Limit Theorem for blurring and nonblurring processes. Then, they are applied to the estimation of location parameter. Simulation studies show that the location estimation based on the convergence point of blurring process is more robust and often more efficient than that of nonblurring process.