Metropolis Scheme Research Articles

Bayesian reconstruction of surface shape from phaseless scattered acoustic data.

The recovery of the properties or geometry of a rough surface from scattered sound is of interest in many applications, including medicine, water engineering, or structural health monitoring. Existing approaches to reconstruct the roughness profile of a scattering surface based on wave scattering have no intrinsic way of predicting the uncertainty of the reconstruction. In an attempt to recover this uncertainty, a Bayesian framework, and more explicitly, an adaptive Metropolis scheme, is used to infer the properties of a rough surface, parameterised as a superposition of sinusoidal components. The Kirchhoff approximation is used in the present work as the underlying model of wave scattering, and is constrained by the assumption of surface smoothness. This implies a validity region in the parameter space, which is incorporated in the Bayesian formulation, making the resulting method physics informed compared to data-based approaches. For a three-parameter sinusoidal surface and a rough surface with a random roughness profile, physical experiments were conducted to collect scattered field data. The models were then tested on the experimental data. The recovery offers insight of the Bayesian approach results expressed in terms of confidence intervals, and could be used as a method to identify uncertainty.

Metropolis Monte Carlo sampling: convergence, localization transition and optimality

Among random sampling methods, Markov chain Monte Carlo (MC) algorithms are foremost. Using a combination of analytical and numerical approaches, we study their convergence properties toward the steady state, within a random walk Metropolis scheme. Analyzing the relaxation properties of some model algorithms sufficiently simple to enable analytic progress, we show that the deviations from the target steady-state distribution can feature a localization transition as a function of the characteristic length of the attempted jumps defining the random walk. While the iteration of the MC algorithm converges to equilibrium for all choices of jump parameters, the localization transition changes drastically the asymptotic shape of the difference between the probability distribution reached after a finite number of steps of the algorithm and the target equilibrium distribution. We argue that the relaxation before and after the localization transition is respectively limited by diffusion and rejection rates.

Group Importance Sampling for particle filtering and MCMC

Bayesian methods and their implementations by means of sophisticated Monte Carlo techniques have become very popular in signal processing over the last years. Importance Sampling (IS) is a well-known Monte Carlo technique that approximates integrals involving a posterior distribution by means of weighted samples. In this work, we study the assignation of a single weighted sample which compresses the information contained in a population of weighted samples. Part of the theory that we present as Group Importance Sampling (GIS) has been employed implicitly in different works in the literature. The provided analysis yields several theoretical and practical consequences. For instance, we discuss the application of GIS into the Sequential Importance Resampling framework and show that Independent Multiple Try Metropolis schemes can be interpreted as a standard Metropolis–Hastings algorithm, following the GIS approach. We also introduce two novel Markov Chain Monte Carlo (MCMC) techniques based on GIS. The first one, named Group Metropolis Sampling method, produces a Markov chain of sets of weighted samples. All these sets are then employed for obtaining a unique global estimator. The second one is the Distributed Particle Metropolis–Hastings technique, where different parallel particle filters are jointly used to drive an MCMC algorithm. Different resampled trajectories are compared and then tested with a proper acceptance probability. The novel schemes are tested in different numerical experiments such as learning the hyperparameters of Gaussian Processes, two localization problems in a wireless sensor network (with synthetic and real data) and the tracking of vegetation parameters given satellite observations, where they are compared with several benchmark Monte Carlo techniques. Three illustrative Matlab demos are also provided.

Bayesian calibration of terrestrial ecosystem models: a study of advanced Markov chain Monte Carlo methods

Abstract. Calibration of terrestrial ecosystem models is important but challenging. Bayesian inference implemented by Markov chain Monte Carlo (MCMC) sampling provides a comprehensive framework to estimate model parameters and associated uncertainties using their posterior distributions. The effectiveness and efficiency of the method strongly depend on the MCMC algorithm used. In this work, a differential evolution adaptive Metropolis (DREAM) algorithm is used to estimate posterior distributions of 21 parameters for the data assimilation linked ecosystem carbon (DALEC) model using 14 years of daily net ecosystem exchange data collected at the Harvard Forest Environmental Measurement Site eddy-flux tower. The calibration of DREAM results in a better model fit and predictive performance compared to the popular adaptive Metropolis (AM) scheme. Moreover, DREAM indicates that two parameters controlling autumn phenology have multiple modes in their posterior distributions while AM only identifies one mode. The application suggests that DREAM is very suitable to calibrate complex terrestrial ecosystem models, where the uncertain parameter size is usually large and existence of local optima is always a concern. In addition, this effort justifies the assumptions of the error model used in Bayesian calibration according to the residual analysis. The result indicates that a heteroscedastic, correlated, Gaussian error model is appropriate for the problem, and the consequent constructed likelihood function can alleviate the underestimation of parameter uncertainty that is usually caused by using uncorrelated error models.

A Spatial Target Function for Metropolis Photon Tracing

The human visual system is sensitive to relative differences in luminance, but light transport simulation algorithms based on Metropolis sampling often result in a highly nonuniform relative error distribution over the rendered image. Although this issue has previously been addressed in the context of the Metropolis light transport algorithm, our work focuses on Metropolis photon tracing. We present a new target function (TF) for Metropolis photon tracing that ensures good stratification of photons leading to pixel estimates with equalized relative error. We develop a hierarchical scheme for progressive construction of the TF from paths sampled during rendering. In addition to the approach taken in previous work, where the TF is defined in the image plane, ours can be associated with compact spatial regions. This allows us to take advantage of illumination coherence to more robustly estimate the TF while adapting to geometry discontinuities. To sample from this TF, we design a new replica exchange Metropolis scheme. We apply our algorithm in progressive photon mapping and show that it often outperforms alternative approaches in terms of image quality by a large margin.

A spatial target function for metropolis photon tracing

The human visual system is sensitive to relative differences in luminance, but light transport simulation algorithms based on Metropolis sampling often result in a highly nonuniform relative error distribution over the rendered image. Although this issue has previously been addressed in the context of the Metropolis light transport algorithm, our work focuses on Metropolis photon tracing. We present a new target function (TF) for Metropolis photon tracing that ensures good stratification of photons leading to pixel estimates with equalized relative error. We develop a hierarchical scheme for progressive construction of the TF from paths sampled during rendering. In addition to the approach taken in previous work, where the TF is defined in the image plane, ours can be associated with compact spatial regions. This allows us to take advantage of illumination coherence to more robustly estimate the TF while adapting to geometry discontinuities. To sample from this TF, we design a new replica exchange Metropolis scheme. We apply our algorithm in progressive photon mapping and show that it often outperforms alternative approaches in terms of image quality by a large margin.

A method to reduce the rejection rate in Monte Carlo Markov chains

We present a method for Monte Carlo sampling on systems with discrete variables (focusing in the Ising case), introducing a prior on the candidate moves in a Metropolis–Hastings scheme which can significantly reduce the rejection rate, called the reduced-rejection-rate (RRR) method. The method employs same probability distribution for the choice of the moves as rejection-free schemes such as the method proposed by Bortz, Kalos and Lebowitz (BKL) (1975 J. Comput. Phys. 17 10–8); however, it uses it as a prior in an otherwise standard Metropolis scheme: it is thus not fully rejection-free, but in a wide range of scenarios it is nearly so. This allows to extend the method to cases for which rejection-free schemes become inefficient, in particular when the graph connectivity is not sparse, but the energy can nevertheless be expressed as a sum of two components, one of which is computed on a sparse graph and dominates the measure. As examples of such instances, we demonstrate that the method yields excellent results when performing Monte Carlo simulations of quantum spin models in presence of a transverse field in the Suzuki–Trotter formalism, and when exploring the so-called robust ensemble which was recently introduced in Baldassi et al (2016 Proc. Natl Acad. Sci. 113 E7655–62). Our code for the Ising case is publicly available (RRR Monte Carlo code https://github.com/carlobaldassi/RRRMC.jl), and extensible to user-defined models: it provides efficient implementations of standard Metropolis, the RRR method, the BKL method (extended to the case of continuous energy specra), and the waiting time method by Dall and Sibani (2001 Comput. Phys. Commun. 141 260–7).

A Hybrid Adaptive MCMC Algorithm in Function Spaces

The preconditioned Crank--Nicolson (pCN) method is a Markov chain Monte Carlo (MCMC) scheme, specifically designed to perform Bayesian inferences in function spaces. Unlike many standard MCMC algorithms, the pCN method can preserve the sampling efficiency under the mesh refinement, a property referred to as being dimension independent. In this work we consider an adaptive strategy to further improve the efficiency of pCN. In particular we develop a hybrid adaptive MCMC method: the algorithm performs an adaptive Metropolis scheme in a chosen finite dimensional subspace and a standard pCN algorithm in the complement space of the chosen subspace. We show that the proposed algorithm satisfies certain important ergodicity conditions. Finally with numerical examples we demonstrate that the proposed method has competitive performance with existing adaptive algorithms.

A Spatial Target Function for Metropolis Photon Tracing

The human visual system is sensitive to relative differences in luminance, but light transport simulation algorithms based on Metropolis sampling often result in a highly nonuniform relative error distribution over the rendered image. Although this issue has previously been addressed in the context of the Metropolis light transport algorithm, our work focuses on Metropolis photon tracing. We present a new target function (TF) for Metropolis photon tracing that ensures good stratification of photons leading to pixel estimates with equalized relative error. We develop a hierarchical scheme for progressive construction of the TF from paths sampled during rendering. In addition to the approach taken in previous work, where the TF is defined in the image plane, ours can be associated with compact spatial regions. This allows us to take advantage of illumination coherence to more robustly estimate the TF while adapting to geometry discontinuities. To sample from this TF, we design a new replica exchange Metropolis scheme. We apply our algorithm in progressive photon mapping and show that it often outperforms alternative approaches in terms of image quality by a large margin.

Formation of carbon Cottrell atmospheres and their effect on the stress field around an edge dislocation

Biased Metropolis Monte-Carlo is employed to build a carbon Cottrell atmosphere around an edge dislocation. A novel method involving only local minimisation during the Metropolis scheme allows to deal with a million-atom system within a reasonable computation time. In this study, it is discovered that (i) the carbon atoms occupy interstitial sites not only in the traction zone, but also in the compression zone; (ii) local carbon concentrations of approximately 10at% are in good agreement with experimental values; (iii) the saturation of the Cottrell atmosphere does not originate from the stress-field around the atmosphere but from repulsive carbon-carbon interactions only.

A Metropolis scheme for Monte Carlo methods for the solution of particle population balance

A Metropolis scheme that can be used by various Monte Carlo methods for solving particle population balance equations has been proposed and verified. A detailed derivation of the scheme is presented. The computational accuracy of the proposed scheme is tested under various coagulation kernels. The computational performance of the scheme is also measured and compared to that of Inverse and acceptance–rejection scheme.

Fast quantum Monte Carlo on a GPU

We present a scheme for the parallelization of quantum Monte Carlo method on graphical processing units, focusing on variational Monte Carlo simulation of bosonic systems. We use asynchronous execution schemes with shared memory persistence, and obtain an excellent utilization of the accelerator. The CUDA code is provided along with a package that simulates liquid helium-4. The program was benchmarked on several models of Nvidia GPU, including Fermi GTX560 and M2090, and the Kepler architecture K20 GPU. Special optimization was developed for the Kepler cards, including placement of data structures in the register space of the Kepler GPUs. Kepler-specific optimization is discussed. Program SummaryProgram title: QLCatalogue identifier: AEUP_v1_0Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEUP_v1_0.htmlProgram obtainable from: CPC Program Library, Queen’s University, Belfast, N. IrelandLicensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.htmlNo. of lines in distributed program, including test data, etc.: 40170No. of bytes in distributed program, including test data, etc.: 1223080Distribution format: tar.gzProgramming language: CUDA-C, C, Fortran.Computer: Any computer with a CUDA-enabled GPU.Operating system: Linux.RAM: Typical execution uses as much RAM as is available on the GPU; usually between 1 GB and 12 GB. Minimal requirement is 1 MB.Classification: 4.12, 7.7.Nature of problem: QL package executes variational Monte Carlo for liquid helium-4 with Aziz II interaction potential and a Jastrow pair product trial wavefunction. Sampling is performed with a Metropolis scheme applied to single-particle updates. With minimal changes, the package can be applied to other bosonic fluids, given a pairwise interaction potential and a wavefunction in the form of a product of one- and two-body correlation factors.Solution method: The program is parallelized for execution with Nvidia GPU. By design, the generation of new configurations is performed with shared memory persistence and the asynchronous execution allows for the CPU load masking.Restrictions: Code is limited to variational Monte Carlo. Due to the limitation of the shared memory of GPU, only systems under 2000 particles can be treated on the Fermi generation cards, and up to 10000 on Kepler cards.Running time: Because of the statistical nature of Monte Carlo calculations, computations may be chained indefinitely to improve statistical accuracy. As an example, using the QL package, the energy of a liquid helium system with 1952 atoms can be computed to within 1 mK per atom in less than 20 min. This corresponds to the relative error of 10−4. It is unlikely that a higher accuracy may be needed.

A conservative and a hybrid early rejection schemes for accelerating Monte Carlo molecular simulation

Molecular simulation could provide detailed description of fluid systems when compared to experimental techniques. They can also replace equations of state; however, molecular simulation usually costs considerable computational efforts. Several techniques have been developed to overcome such high computational costs. In this paper, two early rejection schemes, a conservative and a hybrid one, are introduced. In these two methods, undesired configurations generated by the Monte Carlo trials are rejected earlier than it would when using conventional algorithms. The methods are tested for structureless single-component Lennard–Jones particles in both canonical and NVT-Gibbs ensembles. The computational time reduction for both ensembles is observed at a wide range of thermodynamic conditions. Results show that computational time savings are directly proportional to the rejection rate of Monte Carlo trials. The proposed conservative scheme has shown to be successful in saving up to 40% of the computational time in the canonical ensemble and up to 30% in the NVT-Gibbs ensemble when compared to standard algorithms. In addition, it preserves the exact Markov chains produced by the Metropolis scheme. Further enhancement for NVT-Gibbs ensemble is achieved by combining this technique with the bond formation early rejection one. The hybrid method achieves more than 50% saving of the central processing unit (CPU) time.

On the flexibility of the design of multiple try Metropolis schemes

The Multiple Try Metropolis (MTM) method is a generalization of the classical Metropolis-Hastings algorithm in which the next state of the chain is chosen among a set of samples, according to normalized weights. In the literature, several extensions have been proposed. In this work, we show and remark upon the flexibility of the design of MTM-type methods, fulfilling the detailed balance condition. We discuss several possibilities and show different numerical results.

A multi-point Metropolis scheme with generic weight functions

The multi-point Metropolis algorithm is an advanced MCMC technique based on drawing several correlated samples at each step and choosing one of them according to some normalized weights. We propose a variation of this technique where the weight functions are not specified, i.e., the analytic form can be chosen arbitrarily. This has the advantage of greater flexibility in the design of high-performance MCMC samplers. We prove that our method fulfills the balance condition, and provide a numerical simulation. We also give new insight into the functionality of different MCMC algorithms, and the connections between them.

Monte Carle simulation of quantum transport through nanostructures

We describe a numerical scheme of combining Monte Carlo procedure and quantum scattering theory to simulate electron transport processes through nanostructures. The transport of electrons through a nanostructure is a highly nontrivial nonequilibrium process in which we should consider the interplay of (i) complicated many-body quantum states in nanostructure, (ii) thermal relaxation processes keeping the leads (electron reservoirs) in local equilibrium, (iii) the coupling between the leads and the nanostructure, and (iv) the bias causing nonequilibrium, current, and evolution of quantum states in the nanostructure. Considering the quantum coherence within the nanostructure, we include the degrees of freedom of the nanostructure and a single tunneling electron and solve the Schrödinger equation for the many-body states to obtain the scattering matrix in the Fock space from which both the transmission of the electron and the variation of the states in nanostructure can be full quantum-mechanically calculated. The transport is investigated by the Monte Carlo simulation of successive scattering events of single electrons which are sampled with the Metropolis scheme governed by the scattering probabilities, the thermal statistics in the leads, and the applied bias. By this way from a given initial nanostructure state we can calculate the time evolutions of the current and the internal state. As examples we investigate the transmission of electrons through a two-level system. It is shown that the proposed method can properly deal with the inelastic effects in transport processes.

Simulation of classical thermal states on a quantum computer: A transfer-matrix approach

We present a hybrid quantum-classical algorithm to simulate thermal states of classical Hamiltonians on a quantum computer. Our scheme employs a sequence of locally controlled rotations, building up the desired state by adding qubits one at a time. We identified a class of classical models for which our method is efficient and avoids potential exponential overheads encountered by Grover-like or quantum Metropolis schemes. Our algorithm also gives an exponential advantage for two-dimensional Ising models with magnetic field on a square lattice, compared with the previously known Zalka's algorithm.

Bayesian analysis of a dynamical model for the spread of the Usutu virus

The Usutu virus is an arbovirus transmitted by mosquitoes and causing disease in birds. The virus was detected in Austria for the first time in 2001, while a major outbreak occurred in 2003. Rubel et al. (2008) developed a nine-compartment deterministic SEIR model to explain the spread of the disease. We extended this to a hierarchical Bayes model assuming random variation in temperature data, in reproduction data of birds, and in the number of birds found to be infected. The model was implemented in R, combined with the FORTRAN subroutine for the original deterministic model. Analysis was made by MCMC using a random walk Metropolis scheme. Posterior means, medians, and credible intervals were calculated for the parameters. The hierarchical Bayes approach proved to be fruitful in extending the deterministic model into a stochastic one. It allowed for Bayesian point and interval estimation and quantification of uncertainty of predictions. The analysis revealed that some model parameters were not identifiable; therefore we kept constant some of them and analyzed others conditional on them. Identifiability problems are common in models aiming to mirror the mechanism of the process, since parameters with natural interpretation are likely to exhibit interrelationships. This study illustrated that Bayesian modeling combined with conditional analysis may help in those cases. Its application to the Usutu model improved model fit and revealed the structure of interdependencies between model parameters: it demonstrated that determining some of them experimentally would enable estimation of the others, except one of them, from available data.

Application of Biased Metropolis Algorithms: From protons to proteins

We show that sampling with a biased Metropolis scheme is essentially equivalent to using the heatbath algorithm. However, the biased Metropolis method can also be applied when an efficient heatbath algorithm does not exist. This is first illustrated with an example from high energy physics (lattice gauge theory simulations). We then illustrate the Rugged Metropolis method, which is based on a similar biased updating scheme, but aims at very different applications. The goal of such applications is to locate the most likely configurations in a rugged free energy landscape, which is most relevant for simulations of biomolecules.

Calculation of the surface tension from Monte Carlo simulations: Does the model impact on the finite-size effects?

We report two-phase Monte Carlo simulations of the liquid-vapor interface of the Lennard-Jones (LJ) fluids in order to study the impact of the methodology used for the energy calculation on the oscillatory behavior of the surface tension with the system sizes. The surface tension values are illustrated through the LJ parameters of methane. The first methodology uses a standard truncated LJ potential, the second one adds a long range correction (LRC) contribution to the energy into the Metropolis scheme, and the third one uses a LJ potential modified by a polynomial function in order to remove the discontinuities at the cutoff distance. The surface tension is calculated from the mechanical and thermodynamic routes and the LRCs to the surface tension are systematically calculated from appropriate expressions within these definitions. The oscillatory behavior has been studied as a function of the size of the interfacial area and of the length of the dimension perpendicular to the surface. We show that the methodology has an important effect on the oscillatory variation in the surface tension with the system size. This oscillatory variation in the surface tension with the system size is investigated through its intrinsic and LRC contributions. We complete this work by studying the dependence of the surface tension with respect to the cutoff distance when the LRC part to the energy is considered into the Metropolis scheme.