Bayesian model comparison with un-normalised likelihoods

Models for which the likelihood function can be evaluated only up to a parameter-dependent unknown normalizing constant, such as Markov random field models, are used widely in computer science, statistical physics, spatial statistics, and network analysis. However, Bayesian analysis of these models using standard Monte Carlo methods is not possible due to the intractability of their likelihood functions. Several methods that permit exact, or close to exact, simulation from the posterior distribution have recently been developed. However, estimating the evidence and Bayes’ factors for these models remains challenging in general. This paper describes new random weight importance sampling and sequential Monte Carlo methods for estimating BFs that use simulation to circumvent the evaluation of the intractable likelihood, and compares them to existing methods. In some cases we observe an advantage in the use of biased weight estimates. An initial investigation into the theoretical and empirical properties of this class of methods is presented. Some support for the use of biased estimates is presented, but we advocate caution in the use of such estimates.

Segmentação de dados de perfilamento a laser em áreas urbanas utilizando uma abordagem Bayesiana

In this paper is presented a region-based methodology for Digital Elevation Model segmentation obtained from laser scanning data. The methodology is based on two sequential techniques, i.e., a recursive splitting technique using the quad tree structure followed by a region merging technique using the Markov Random Field model. The recursive splitting technique starts splitting the Digital Elevation Model into homogeneous regions. However, due to slight height differences in the Digital Elevation Model, region fragmentation can be relatively high. In order to minimize the fragmentation, a region merging technique based on the Markov Random Field model is applied to the previously segmented data. The resulting regions are firstly structured by using the so-called Region Adjacency Graph. Each node of the Region Adjacency Graph represents a region of the Digital Elevation Model segmented and two nodes have connectivity between them if corresponding regions share a common boundary. Next it is assumed that the random variable related to each node, follows the Markov Random Field model. This hypothesis allows the derivation of the posteriori probability distribution function whose solution is obtained by the Maximum a Posteriori estimation. Regions presenting high probability of similarity are merged. Experiments carried out with laser scanning data showed that the methodology allows to separate the objects in the Digital Elevation Model with a low amount of fragmentation.

Building roof contour extraction from LiDAR data

This paper proposes a method for the automatic extraction of building roof contours from a LiDAR-derived digital surface model (DSM). The method is based on two steps. First, to detect aboveground objects (buildings, trees, etc.), the DSM is segmented through a recursive splitting technique followed by a region merging process. Vectorization and polygonization are used to obtain polyline representations of the detected aboveground objects. Second, building roof contours are identified from among the aboveground objects by optimizing a Markov-random-field-based energy function that embodies roof contour attributes and spatial constraints. Preliminary results have shown that the proposed methodology works properly.

Simulated annealing for building roof contours identification from lidar data

This paper proposes a method by simulated annealing for building roof contours identification from LiDAR-derived digital elevation model. Our method is based on the concept of first extracting aboveground objects and then identifying those objects that are building roof contours. First, to detect aboveground objects (buildings, trees, etc.), the digital elevation model is segmented through a recursive splitting technique followed by a region merging process. Vectorization and polygonization are used to obtain polyline representations of the detected aboveground objects. Second, building roof contours are identified from among the aboveground objects by optimizing a Markov-random-field-based energy function that embodies roof contour attributes and spatial constraints. The solution of this function is a polygon set corresponding to building roof contours and is found by using a minimization technique, like the Simulated Annealing algorithm. Experiments carried out with laser scanning digital elevation model showed that the methodology works properly, as it provides roof contour information with approximately 90% shape accuracy and no verified false positives.

Improved cerebellar tissue classification on magnetic resonance images of brain.

PURPOSE: To develop and implement a method for improved cerebellar tissue classification on the MRI of brain by automatically isolating the cerebellum prior to segmentation. MATERIALS AND METHODS: Dual fast spin echo (FSE) and fluid attenuation inversion recovery (FLAIR) images were acquired on 18 normal volunteers on a 3 T Philips scanner. The cerebellum was isolated from the rest of the brain using a symmetric inverse consistent nonlinear registration of individual brain with the parcellated template. The cerebellum was then separated by masking the anatomical image with individual FLAIR images. Tissues in both the cerebellum and rest of the brain were separately classified using hidden Markov random field (HMRF), a parametric method, and then combined to obtain tissue classification of the whole brain. The proposed method for tissue classification on real MR brain images was evaluated subjectively by two experts. The segmentation results on Brainweb images with varying noise and intensity nonuniformity levels were quantitatively compared with the ground truth by computing the Dice similarity indices. RESULTS: The proposed method significantly improved the cerebellar tissue classification on all normal volunteers included in this study without compromising the classification in remaining part of the brain. The average similarity indices for gray matter (GM) and white matter (WM) in the cerebellum are 89.81 (+/-2.34) and 93.04 (+/-2.41), demonstrating excellent performance of the proposed methodology. CONCLUSION: The proposed method significantly improved tissue classification in the cerebellum. The GM was overestimated when segmentation was performed on the whole brain as a single object.

Uniformly reweighted belief propagation for distributed Bayesian hypothesis testing

Belief propagation (BP) is a technique for distributed inference in wireless networks and is often used even when the underlying graphical model contains cycles. In this paper, we propose a uniformly reweighted BP scheme that reduces the impact of cycles by weighting messages by a constant ?edge appearance probability? rho ? 1. We apply this algorithm to distributed binary hypothesis testing problems (e.g., distributed detection) in wireless networks with Markov random field models. We demonstrate that in the considered setting the proposed method outperforms standard BP, while maintaining similar complexity. We then show that the optimal ? can be approximated as a simple function of the average node degree, and can hence be computed in a distributed fashion through a consensus algorithm.

Brain tissue segmentation on Diffusion Weighted Magnetic Resonance data

Introduction Diffusion weighted Imaging (DWI) techniques are able to measure, in vivo and non-invasively, the diffusivity of water molecules inside the human brain. DWI has been applied on cerebral ischemia, brain maturation, epilepsy, multiple sclerosis, etc. [1]. Nowadays, there is a very high availability of these images. DWI allows the identification of brain tissues, so its accurate segmentation is a common initial step for the referred applications. Materials and Methods We present a validation study on automated segmentation of DWI based on the Gaussian mixture and hidden Markov random field models. This methodology is widely solved with iterative conditional modes algorithm, but some studies suggest [2] that graph-cuts (GC) algorithms improve the results when initialization is not close to the final solution. We implemented a segmentation tool integrating ITK with a GC algorithm [3], and a validation software using fuzzy overlap measures [4]. Results Segmentation accuracy of each tool is tested against a gold-standard segmentation obtained from a T1 MPRAGE magnetic resonance image of the same subject, registered to the DWI space. The proposed software shows meaningful improvements by using the GC energy minimization approach on DTI and DSI (Diffusion Spectrum Imaging) data. Conclusions The brain tissues segmentation on DWI is a fundamental step on many applications. Accuracy and robustness improvements are achieved with the proposed software, with high impact on the application’s final result.

MBIS: Multivariate Bayesian Image Segmentation Tool

We present MBIS (Multivariate Bayesian Image Segmentation tool), a clustering tool based on the mixture of multivariate normal distributions model. MBIS supports multi-channel bias field correction based on a B-spline model. A second methodological novelty is the inclusion of graph-cuts optimization for the stationary anisotropic hidden Markov random field model. Along with MBIS, we release an evaluation framework that contains three different experiments on multi-site data. We first validate the accuracy of segmentation and the estimated bias field for each channel. MBIS outperforms a widely used segmentation tool in a cross-comparison evaluation. The second experiment demonstrates the robustness of results on atlas-free segmentation of two image sets from scan-rescan protocols on 21 healthy subjects. Multivariate segmentation is more replicable than the monospectral counterpart on T1-weighted images. Finally, we provide a third experiment to illustrate how MBIS can be used in a large-scale study of tissue volume change with increasing age in 584 healthy subjects. This last result is meaningful as multivariate segmentation performs robustly without the need for prior knowledge.

Speeding up the EM algorithm for mixture model-based segmentation of magnetic resonance images

Mixture models implemented via the expectation-maximization (EM) algorithm are being increasingly used in a wide range of problems in pattern recognition such as image segmentation. However, the EM algorithm requires considerable computational time in its application to huge data sets such as a three-dimensional magnetic resonance (MR) image of over 10 million voxels. Recently, it was shown that a sparse, incremental version of the EM algorithm could improve its rate of convergence. In this paper, we show how this modified EM algorithm can be speeded up further by adopting a multiresolution kd-tree structure in performing the E-step. The proposed algorithm outperforms some other variants of the EM algorithm for segmenting MR images of the human brain. (C) 2004 Pattern Recognition Society. Published by Elsevier Ltd. All rights reserved.

A generative model for separating illumination and reflectance from images

It is well known that even slight changes in nonuniform illumination lead to a large image variability and are crucial for many visual tasks. This paper presents a new ICA related probabilistic model where the number of sources exceeds the number of sensors to perform an image segmentation and illumination removal, simultaneously. We model illumination and reflectance in log space by a generalized autoregressive process and Hidden Gaussian Markov random field, respectively. The model ability to deal with segmentation of illuminated images is compared with a Canny edge detector and homomorphic filtering. We apply the model to two problems: synthetic image segmentation and sea surface pollution detection from intensity images.

A Bayesian Spatial Mixture Model for FMRI Analysis

We develop, implement and study a new Bayesian spatial mixture model (BSMM). The proposed BSMM allows for spatial structure in the binary activation indicators through a latent thresholded Gaussian Markov random field. We develop a Gibbs (MCMC) sampler to perform posterior inference on the model parameters, which then allows us to assess the posterior probabilities of activation for each voxel. One purpose of this article is to compare the HJ model and the BSMM in terms of receiver operating characteristics (ROC) curves. Also we consider the accuracy of the spatial mixture model and the BSMM for estimation of the size of the activation region in terms of bias, variance and mean squared error. We perform a simulation study to examine the aforementioned characteristics under a variety of configurations of spatial mixture model and BSMM both as the size of the region changes and as the magnitude of activation changes.

Probabilistic Models for Scalable Knowledge Graph Construction

In the past decade, systems that extract information from millions of Internet documents have become commonplace. Knowledge graphs -- structured knowledge bases that describe entities, their attributes and the relationships between them -- are a powerful tool for understanding and organizing this vast amount of information. However, a significant obstacle to knowledge graph construction is the unreliability of the extracted information, due to noise and ambiguity in the underlying data or errors made by the extraction system and the complexity of reasoning about the dependencies between these noisy extractions. My dissertation addresses these challenges by exploiting the interdependencies between facts to improve the quality of the knowledge graph in a scalable framework. I introduce a new approach called knowledge graph identification (KGI), which resolves the entities, attributes and relationships in the knowledge graph by incorporating uncertain extractions from multiple sources, entity co-references, and ontological constraints. I define a probability distribution over possible knowledge graphs and infer the most probable knowledge graph using a combination of probabilistic and logical reasoning. Such probabilistic models are frequently dismissed due to scalability concerns, but my implementation of KGI maintains tractable performance on large problems through the use of hinge-loss Markov random fields, which have a convex inference objective. This allows the inference of large knowledge graphs using 4M facts and 20M ground constraints in 2 hours. To further scale the solution, I develop a distributed approach to the KGI problem which runs in parallel across multiple machines, reducing inference time by 90%. Finally, I extend my model to the streaming setting, where a knowledge graph is continuously updated by incorporating newly extracted facts. I devise a general approach for approximately updating inference in convex probabilistic models, and quantify the approximation error by defining and bounding inference regret for online models. Together, my work retains the attractive features of probabilistic models while providing the scalability necessary for large-scale knowledge graph construction. These models have been applied on a number of real-world knowledge graph projects, including the NELL project at Carnegie Mellon and the Google Knowledge Graph.

Induced random fields in the LiHoxY1-xF4 quantum ising magnet in a transverse magnetic field

The LiHoxY1-xF4 magnetic material in a transverse magnetic field Bxx̂ perpendicular to the Ising spin direction has long been used to study tunable quantum phase transitions in a random disordered system. We show that the Bx-induced magnetization along the x̂ direction, combined with the local random dilution-induced destruction of crystalline symmetries, generates, via the predominant dipolar interactions between Ho3+ ions, random fields along the Ising ẑ direction. This identifies LiHoxY1-xF4 in Bx as a new random field Ising system. The random fields explain the rapid decrease of the critical temperature in the diluted ferromagnetic regime and the smearing of the nonlinear susceptibility at the spin-glass transition with increasing Bx and render the Bx-induced quantum criticality in LiHoxY1-xF4 likely inaccessible.

Influence of Spatial Variability on Pavement Responses Using Latin Hypercube Sampling on Two-Dimensional Random Fields

Although uncertainties in material properties have been addressed in the design of flexible pavements, most current modeling techniques assume that pavement layers are homogeneous. The paper addresses the influence of the spatial variability of the resilient moduli of pavement layers by evaluating the effect of the variance and correlation length on the pavement responses to loading. The integration of the spatially varying log-normal random field with the finite-difference method has been achieved through an exponential autocorrelation function. The variation in the correlation length was found to have a marginal effect on the mean values of the critical strains and a noticeable effect on the standard deviation which decreases with decreases in correlation length. This reduction in the variance arises because of the spatial averaging phenomenon over the softer and stiffer zones generated because of spatial variability. The increase in the mean value of critical strains with decreasing correlation length, although minor, illustrates that pavement performance is adversely affected by the presence of spatially varying layers. The study also confirmed that the higher the variability in the pavement layer moduli, introduced through a higher value of coefficient of variation (COV), the higher the variability in the pavement response. The study concludes that ignoring spatial variability by modeling the pavement layers as homogeneous that have very short correlation lengths can result in the underestimation of the critical strains and thus an inaccurate assessment of the pavement performance. (C) 2014 American Society of Civil Engineers.

Matrix Crack Detection in Composite Plate with Spatially Random Material Properties using Fractal Dimension

Fractal dimension based damage detection method is investigated for a composite plate with random material properties. Composite material shows spatially varying random material properties because of complex manufacturing processes. Matrix cracks are considered as damage in the composite plate. Such cracks are often seen as the initial damage mechanism in composites under fatigue loading and also occur due to low velocity impact. Static deflection of the cantilevered composite plate with uniform loading is calculated using the finite element method. Damage detection is carried out based on sliding window fractal dimension operator using the static deflection. Two dimensional homogeneous Gaussian random field is generated using Karhunen-Loeve (KL) expansion to represent the spatial variation of composite material property. The robustness of fractal dimension based damage detection method is demonstrated considering the composite material properties as a two dimensional random field.