Ajay Jasra

Markov Chain Monte Carlo Methods and the Label Switching Problem in Bayesian Mixture Modeling

In the past ten years there has been a dramatic increase of in terest in the Bayesian analysis of finite mixture models. This is primarily because of the emergence of Markov chain Monte Carlo (MCMC) methods. While MCMC provides a convenient way to draw inference from compli cated statistical models, there are many, perhaps underappreciated, problems associated with the MCMC analysis of mixtures. The problems are mainly caused by the nonidentifiability of the components under symmetric priors, which leads to so-called label switching in the MCMC output. This means that ergodic averages of component specific quantities will be identical and thus useless for inference. We review the solutions to the label switching problem, such as artificial identifiability constraints, relabelling algorithms and label invariant loss functions. We also review various MCMC sampling schemes that have been suggested for mixture models and discuss posterior sensitivity to prior specification.

On adaptive resampling strategies for sequential Monte Carlo methods

Sequential Monte Carlo (SMC) methods are a general class of techniques to sample approximately from any sequence of probability distributions. These distributions are approximated by a cloud of weighted samples which are propagated over time using a combination of importance sampling and resampling steps. This article is concerned with the convergence analysis of a class of SMC methods where the times at which resampling occurs are computed on-line using criteria such as the effective sample size. This is a popular approach amongst practitioners but there are very few convergence results available for these methods. It is shown here that these SMC algorithms correspond to a particle approximation of a Feynman-Kac flow of measures on adaptive excursion spaces. By combining a non-linear distribution flow analysis to an original coupling technique, we obtain functional central limit theorems and uniform exponential concentration estimates for these algorithms. The original exponential concentration theorems presented in this study significantly improve previous concentration estimates obtained for SMC algorithms.

On the stability of sequential Monte Carlo methods in high dimensions

We investigate the stability of a Sequential Monte Carlo (SMC) method applied to the problem of sampling from a target distribution on Rd for large d. It is well known [Bengtsson, Bickel and Li, In Probability and Statistics: Essays in Honor of David A. Freedman, D. Nolan and T. Speed, eds. (2008) 316–334 IMS; see also Pushing the Limits of Contemporary Statistics (2008) 318–32 9 IMS, Mon. Weather Rev. (2009) 136 (2009) 4629–4640] that using a single importance sampling step, one produces an approximation for the target that deteriorates as the dimension d increases, unless the number of Monte Carlo samples N increases at an exponential rate in d. We show that this degeneracy can be avoided by introducing a sequence of artificial targets, starting from a “simple” density and moving to the one of interest, using an SMC method to sample from the sequence; see, for example, Chopin [Biometrika 89 (2002) 539–551]; see also [J. R. Stat. Soc. Ser. B Stat. Methodol. 68 (2006) 411–436, Phys. Rev. Lett. 78 (1997) 2690–2693, Stat. Comput. 11 (2001) 125–139]. Using this class of SMC methods with a fixed number of samples, one can produce an approximation for which the effective sample size (ESS) converges to a random variable eN as d → ∞ with 1 < eN < N. The convergence is achieved with a computational cost proportional to Nd2. If eN � N, we can raise its value by introducing a number of resampling steps, say m (where m is independent of d). In this case, the ESS converges to a random variable eN,m as d → ∞ and limm→∞ eN,m = N. Also, we show that the Monte Carlo error for estimating a fixed-dimensional marginal expectation is of order √ 1 N uniformly in d. The results imply that, in high dimensions, SMC algorithms can efficiently control the variability of the importance sampling weights and estimate fixed-dimensional marginals at a cost which is less than exponential in d and indicate that resampling leads to a reduction in the Monte Carlo error and increase in the ESS. All of our analysis is made under the assumption that the target density is i.i.d.

Robust Finite-Time Control of Switched Linear Systems and Application to a Class of Servomechanism Systems

This paper investigates finite-time (FT) stability and stabilization problems for a class of switched linear systems with polytopic uncertainties. Both stable and unstable subsystems are considered to coexist in the system, and a new concept of extended FT stability is proposed as the first attempt. A stability criterion is first established, where the admissible maximum switching number is obtained while ensuring extended FT stability of switched linear systems with time-varying delays under a given maximum ratio between the running time of unstable subsystems and the running time of stable subsystems. Sufficient conditions on the existence of desired memory state-feedback controllers are then developed. A numerical example and a class of servomechanism systems are given, respectively, to illustrate the effectiveness and validity of the developed techniques with time-varying delays and without time delay.

Interacting sequential Monte Carlo samplers for trans-dimensional simulation

The methodology of interacting sequential Monte Carlo (SMC) samplers is introduced. SMC samplers are methods for sampling from a sequence of densities on a common measurable space using a combination of Markov chain Monte Carlo (MCMC) and sequential importance sampling/resampling (SIR) methodology. One of the main problems with SMC samplers when simulating from trans-dimensional, multimodal static targets is that transition kernels do not mix which leads to low particle diversity. In such situations poor Monte Carlo estimates may be derived. To deal with this problem an interacting SMC approach for static inference is introduced. The method proceeds by running SMC samplers in parallel on, initially, different regions of the state space and then moving the corresponding samples onto the entire state space. Once the samplers reach a common space the samplers are combined and allowed to interact. The method is intended to increase the diversity of the population of samples. It is established that interacting SMC admit a Feynman-Kac representation; this provides a framework for the convergence results that are developed. In addition, the methodology is demonstrated on a trans-dimensional inference problem in Bayesian mixture modelling and also, using adaptive methods, a mixture modelling problem in population genetics.

DECODE: a new method for discovering clusters of different densities in spatial data

When clusters with different densities and noise lie in a spatial point set, the major obstacle to classifying these data is the determination of the thresholds for classification, which may form a series of bins for allocating each point to different clusters. Much of the previous work has adopted a model-based approach, but is either incapable of estimating the thresholds in an automatic way, or limited to only two point processes, i.e. noise and clusters with the same density. In this paper, we present a new density-based cluster method (DECODE), in which a spatial data set is presumed to consist of different point processes and clusters with different densities belong to different point processes. DECODE is based upon a reversible jump Markov Chain Monte Carlo (MCMC) strategy and divided into three steps. The first step is to map each point in the data to its mth nearest distance, which is referred to as the distance between a point and its mth nearest neighbor. In the second step, classification thresholds are determined via a reversible jump MCMC strategy. In the third step, clusters are formed by spatially connecting the points whose mth nearest distances fall into a particular bin defined by the thresholds. Four experiments, including two simulated data sets and two seismic data sets, are used to evaluate the algorithm. Results on simulated data show that our approach is capable of discovering the clusters automatically. Results on seismic data suggest that the clustered earthquakes, identified by DECODE, either imply the epicenters of forthcoming strong earthquakes or indicate the areas with the most intensive seismicity, this is consistent with the tectonic states and estimated stress distribution in the associated areas. The comparison between DECODE and other state-of-the-art methods, such as DBSCAN, OPTICS and Wavelet Cluster, illustrates the contribution of our approach: although DECODE can be computationally expensive, it is capable of identifying the number of point processes and simultaneously estimating the classification thresholds with little prior knowledge.

Extended finite-time H∞ control for uncertain switched linear neutral systems with time-varying delays

This paper is concerned with the finite-time H ∞ control problem for a class of uncertain switched linear neutral systems with time-varying delays. By fully exploiting the mode-dependency of the proposed Lyapunov-Krasovskii functionals, the considered systems could be further investigated with stable and unstable subsystems being concurrently contained. Sufficient criteria are first derived for ensuring finite-time boundedness of the underlying system with a given maximum ratio of activation time between unstable subsystems and stable ones. Then the H ∞ performance analysis and the memory state-feedback controller design are carried out for a given performance index. Finally, two numerical examples are presented to illustrate the potential and advantages of the developed findings.

Error bounds and normalising constants for sequential Monte Carlo samplers in high dimensions

In this paper we develop a collection of results associated to the analysis of the sequential Monte Carlo (SMC) samplers algorithm, in the context of high-dimensional independent and identically distributed target probabilities. The SMC samplers algorithm can be designed to sample from a single probability distribution, using Monte Carlo to approximate expectations with respect to this law. Given a target density in d dimensions our results are concerned with d → ∞, while the number of Monte Carlo samples, N, remains fixed. We deduce an explicit bound on the Monte-Carlo error for estimates derived using the SMC sampler and the exact asymptotic relative -error of the estimate of the normalising constant associated to the target. We also establish marginal propagation of chaos properties of the algorithm. These results are deduced when the cost of the algorithm is O(Nd 2).

Sequential Monte Carlo Methods for High-Dimensional Inverse Problems: A Case Study for the Navier-Stokes Equations ∗

We consider the inverse problem of estimating the initial condition of a partial differential equation, which is observed only through noisy measurements at discrete time intervals. In particular, we focus on the case where Eulerian measurements are obtained from the time and space evolving vector field, whose evolution obeys the two-dimensional Navier--Stokes equations defined on a torus. This context is particularly relevant to the area of numerical weather forecasting and data assimilation. We will adopt a Bayesian formulation resulting from a particular regularization that ensures the problem is well posed. In the context of Monte Carlo--based inference, it is a challenging task to obtain samples from the resulting high-dimensional posterior on the initial condition. In real data assimilation applications it is common for computational methods to invoke the use of heuristics and Gaussian approximations. As a result, the resulting inferences are biased and not well justified in the presence of nonlinea...

On nonlinear Markov chain Monte Carlo

Let