Mixture Model Method Research Articles

Soft sensor design based on phase partition ensemble of LSSVR models for nonlinear batch processes.

Traditional single model based soft sensors may have poor performance on quality prediction for batch processes because of the strong nonlinearity, multiple-phase, and time-varying characteristics. Therefore, a phase partition based ensemble learning framework upon least squares support vector regression (LSSVR) is proposed for soft sensor modeling. Firstly, multiway principal component analysis (MPCA) is employed to handle high-dimensional datasets and extract essential correlation information. Then, different operation phases of the process can be identified by the phase partition strategy based on Gaussian mixture model (GMM) method. Meanwhile, the optimal Gaussian component number is determined by Bayesian information criterion (BIC) technique. Further, multiple localized LSSVR models are constructed to characterize the various dynamic relationships between quality and process variables for local regions, while the grid search (GS) and ten-fold cross-validation methods are introduced to parameter optimization for each local model. Finally, the posterior probability for each test sample with respect to different phases can be estimated by Bayesian inference strategy, and local outputs are integrated to produce the final quality prediction results. Feasibility and superiority of the proposed soft sensor are validated through a case study for penicillin fermentation process. It can achieve satisfactory prediction accuracy and effectively tackle nonlinear and multi-phase modeling problems in chemical and biological processes.

An improved Gaussian mixture modeling algorithm combining foreground matching and short-term stability measure for motion detection

Although deep learning has been successfully applied in image analysis, the conventional Gaussian Mixture Modeling (GMM) method still has great potential for multi-mode motion detection because it does not require the support of specialized hardware such as GPUs and massive training data. Under the framework of GMM, this paper combines foreground matching and short-term stability measure to detect slow-moving objects. Foreground models built and updated using the detected foreground pixels have the priority to match potential foreground in the incoming pixels. Meanwhile, the pixel-level stability is measured to make sure that an integrated foreground is detected when a dynamic foreground process is followed. The combination of foreground matching and short-term stability measure greatly improves GMM’s tolerance to slow-moving objects. The quantitative evaluation demonstrates the effectiveness of the proposed algorithm to robustly detect slow-moving objects under a variety of real environments with distracting motions such as illumination changes.

Dynamic probability modeling-based aircraft structural health monitoring framework under time-varying conditions: Validation in an in-flight test simulated on ground

This paper proposes a dynamic probability modeling-based aircraft Structural Health Monitoring (SHM) framework which provides a modular hierarchical SHM architecture for the development of applicable aircraft SHM techniques. The problem of reliable damage monitoring under time-varying conditions, which is the main application obstacle, is fully considered in the framework by the double probability models combined with the short term and long term dynamic update of the models. To realize the SHM capability of the framework, an adaptive constructing method of Gaussian mixture model is proposed for stable and efficient probability modeling and the probability similarity between dynamically updated models is measured for normalized damage detection. The framework is realized by combining with the guided wave SHM technique and validated in a full-scale aircraft fatigue test which is an in-flight test simulated on ground. The cracks of the right landing gear spar and the left wing panel on the aircraft structure are monitored reliably under the fatigue load conditions.

Detection of Tree Crown from Satellite Imagery using Object Based Image Examination

Detection and delineation of individual tree mainly depends on high resolution satellite images or LiDAR data. Urban green structure, specially urban trees plays a key role in enhancing the life of people. Now a day’s more than half of population is leaving in cities and urban areas. Methods to quantify and monitor trees are not efficient. The traditional methods for forest survey and ground survey are complex because of changes occurs in urban environment. The objective of this research is to extract vegetation using colour based and decision tree method, which can be further sub-classify to obtain area under tree canopy. The results obtained through Object-Based Image Analysis (OBIA) method are also compared with existing Gaussian Mixture Model (GMM) method. The overall accuracy achieved thereby is 93.85% using Decision tree-multiresolution segmentation and 93.31% using Decision tree-GMM method.

Research on classification and influencing factors of metro commuting patterns by combining smart card data and household travel survey data

Smart card data (SCD) provide a new perspective for analysing the long-term spatiotemporal travel characteristics of public transit users. Understanding the commuting patterns provides useful insights for urban traffic management. This study attempts to identify and cluster commuting patterns and explore the influencing factors by combining SCD and traditional household travel survey data (HTSD) in Nanjing, China. First, the authors generate the commuting regularity rules using one-day HTSD. Then, the regular metro commuters are identified in four-week (20-weekday) SCD. Using the clustering method of the Gaussian mixture model, they classify metro commuters in SCD into three commuting pattern groups, namely, classic pattern, off-peak pattern, and long-distance pattern, based on their spatiotemporal characteristics. Next, they identify the corresponding metro commuters of these three groups in HTSD and apply a mixed logit regression model to determine the factors influencing metro commuting patterns from multiple dimensions. The results show that some socioeconomic attributes (e.g. gender, age, annual income, education, and occupation) as well as bus station density, metro lines, transfer mode, and transfer distance significantly impact commuting patterns. The findings can provide valuable information for planners and managers to put forward relevant transport guiding measures for alleviating traffic congestion and improving urban traffic management.

Automatic Corneal Ulcer Segmentation Combining Gaussian Mixture Modeling and Otsu Method.

In this paper, we proposed and validated a novel and accurate pipeline for automatically segmenting flaky corneal ulcer areas from fluorescein staining images. The ulcer area was segmented within the cornea by employing a joint method of Otsu and Gaussian Mixture Modeling (GMM). In the GMM based segmentation, the total number of Gaussians was determined intelligently using an information theory based algorithm. And the fluorescein staining images were processed in the HSV color model rather than the original RGB color model, aiming to improve the segmentation results' robustness and accuracy. In the Otsu based segmentation, the images were processed in the grayscale space with Gamma correction being conducted before the Otsu binarization. Afterwards, morphological operations and median filtering were employed to further improve the Otsu segmentation result. The GMM and Otsu segmentation results were then intersected, for which post-processing was conducted by identifying and filling holes through a fast algorithm using priority queues of pixels. The proposed pipeline has been validated on a total of 150 clinical images. Accurate ulcer segmentation results have been obtained, with the mean Dice Similarity Coefficient (DSC) being 0.88 when comparing the automatic segmentation result with the manually-delineated gold standard. For images in the RGB color space, the mean DSC was 0.83, being much lower than that of the images in the HSV color space.

APOE Effect on Amyloid-β PET Spatial Distribution, Deposition Rate, and Cut-Points.

There are conflicting results regarding how APOE genotype, the strongest genetic risk factor for Alzheimer's disease (AD), influences spatial and longitudinal amyloid-β (Aβ) deposition and its impact on the selection of biomarker cut-points. In our study, we sought to determine the impact of APOE genotype on cross-sectional and longitudinal florbetapir positron emission tomography (PET) amyloid measures and its impact in classification of patients and interpretation of clinical cohort results. We included 1,019 and 1,072 Alzheimer's Disease Neuroimaging Initiative participants with cerebrospinal fluid Aβ1-42 and florbetapir PET values, respectively. 623 of these subjects had a second florbetapir PET scans two years after the baseline visit. We evaluated the effect of APOE genotype on Aβ distribution pattern, pathological biomarker cut-points, cross-sectional clinical associations with Aβ load, and longitudinal Aβ deposition rate measured using florbetapir PET scans. 1) APOEɛ4 genotype influences brain amyloid deposition pattern; 2) APOEɛ4 genotype does not modify Aβ biomarker cut-points estimated using unsupervised mixture modeling methods if white matter and brainstem references are used (but not when cerebellum is used as a reference); 3) findings of large differences in Aβ biomarker value differences based on APOE genotype are due to increased probability of having AD neuropathology and are most significant in mild cognitive impairment subjects; and 4) APOE genotype and age (but not gender) were associated with increased Aβ deposition rate. APOEɛ4 carrier status affects rate and location of brain Aβ deposition but does not affect choice of biomarker cut-points if adequate references are selected for florbetapir PET processing.

Robot Learning of Assistive Manipulation Tasks by Demonstration via Head Gesture-based Interface.

Assistive robotic manipulators have the potential to support the lives of people suffering from severe motor impairments. They can support individuals with disabilities to independently perform daily living activities, such as drinking, eating, manipulation tasks, and opening doors. An attractive solution is to enable motor impaired users to teach a robot by providing demonstrations of daily living tasks. The user controls the robot 'manually' with an intuitive human-robot interface to provide demonstration, which is followed by the robot learning of the performed task. However, the control of robotic manipulators by motor impaired individuals is a challenging topic. In this paper, a novel head gesture-based interface for hands-free robot control and a framework for robot learning from demonstration are presented. The head gesture-based interface consists of a camera mounted on the user's hat, which records the changes in the viewed scene due to the head motion. The head gesture recognition is performed using the optical flow for feature extraction and support vector machine for gesture classification. The recognized head gestures are further mapped into robot control commands to perform object manipulation task. The robot learns the demonstrated task by generating the sequence of actions and Gaussian Mixture Model method is used to segment the demonstrated path of the robot's end-effector. During the robotic reproduction of the task, the modified Gaussian Mixture Model and Gaussian Mixture Regression are used to adapt to environmental changes. The proposed framework was evaluated in a real-world assistive robotic scenario in a small study involving 13 participants; 12 able-bodied and one tetraplegic. The presented results demonstrate a potential of the proposed framework to enable severe motor impaired individuals to demonstrate daily living tasks to robotic manipulators.

Estimated Vehicle Density Based on Video Processing Using the Gaussian Mixture Model Method

Congestion is a problem that cannot be decomposed properly, especially at big cities in Indonesia, one of them is in Surabaya. To solve this problem is needed to observe characteristics of the road such as the number of vehicles, vehicle speed and vehicle density. This research took the data in several areas in Surabaya by put the camera in the middle of a crossing bridge. This research aims to determine the density value by comparing traffic volume with vehicle speed. The value of speed and number of vehicles can be known using Gaussian Mixture Model method, where object detection is detected by background substraction process, which is video divided into several frames and in each frame images is divided into two part there are foreground and background. The object will be count 1 when blob analysis detect the foreground. In the theory to know the density value of a traffic flow, with compare the number of vehicles and the speed of vehicle. From the data has been taken, it is known that the number of vehicles is opposite with average speed the vehicle that passes. If average velocity of vehicle greater than the number of vehicle so the value density vehicle getting smaller . so that information can be found that the road is experiencing congestion conditions.

Surface electromyography–based hand movement recognition using the Gaussian mixture model, multilayer perceptron, and AdaBoost method

Human movement is closely linked with muscle activities. Research has indicated that predicting human movements with surface electromyography signals is feasible. However, the classification accura...

Efficient level‐set segmentation model driven by the local GMM and split Bregman method

An efficient level-set model driven by the local Gaussian mixture model (GMM) and split Bregman method is proposed for image segmentation. Firstly, the two intensity fitting functions in the original local binary fitting (LBF) model are pre-estimated by using a local GMM-based intensity distribution estimator before curve evolution. The benefit of this pre-computation strategy is to avoid updating the fitting functions at each step of the curve evolution, and it also overcomes the initialisation problem of common gradient descent-based active contour models, i.e. the level-set function can be initialised as an arbitrary random matrix instead of a signed distance function in the proposed processing framework. Secondly, two processing ideas named global convex segmentation (GCS) method and split Bregman are introduced into the numerical implementation, where the role of GCS is to transform the proposed model into a simplified convex segmentation model, and the purpose of the split Bregman is to quickly output a convergent solution of the newly derived convex segmentation model in an alternate iterative format. Experimental results for synthetic and real images of different modalities with inhomogeneity or homogeneity validate the desired performances of the proposed method in terms of accuracy, robustness, and rapidity.

Handling Device Heterogeneity and Orientation Using Multistage Regression for GMM Based Localization in IoT Networks

Location estimation of heterogeneous smart devices is needed for the Internet of Things (IoT) based location services. Device orientation and heterogeneity are the bottlenecks in accurate location estimation, which are not addressed together in the existing methods. Also, most of the state-of-the-art Received Signal Strength (RSS) based localization methods consider a single Gaussian model instead of a mixture of Gaussians. In this paper, we propose to solve both these issues with a combination of multistage linear regression and Gaussian Mixture Model (GMM) method. Additionally, the proposed method detects the malicious data in the IoT network and estimates the location in case of sensor faults. The performance of the proposed method is tested using Wi-Fi signals in an indoor environment.

Do two parties represent the US? Clustering analysis of US public ideology survey

Recent surveys have shown that an increasing portion of the US public believes the two major US parties adequately represent the US public opinion and think additional parties are needed. However, there are high barriers for third parties in political elections. In this paper, we aim to address two questions: "How well do the two major US parties represent the public's ideology?" and "Does a more-than-two-party system better represent the ideology of the public?". To address these questions, we utilize the American National Election Studies Time series dataset. We perform unsupervised clustering with Gaussian Mixture Model method on this dataset. When clustered into two clusters, we find a large centrist cluster and a small right-wing cluster. The Democratic Party's position (estimated using the mean position of the individuals self-identified with the parties) is similar to that of the centrist cluster, and the Republican Party's position is between the two clusters. We investigate if more than two parties represent the population better by comparing the Akaike Information Criteria for clustering results of the various number of clusters. We find that additional clusters give a better representation of the data, even after penalizing for the additional parameters. This suggests a multiparty system represents of the ideology of the public better.

A fully automatic methodology for MRI brain tumour detection and segmentation

ABSTRACTIn this paper, a complete and fully automatic MRI brain tumour detection and segmentation methodology is presented as an efficient clinical-aided tool using Gaussian mixture model, Fuzzy C-Means, active contour, wavelet transform and entropy segmentation methods. The proposed algorithm is based on two main parts: the skull stripping and tumour auto-detection and segmentation. The first part was evaluated using IBSR, LPBA40 and OASIS databases, and the obtained results show that our proposed method outclasses the best popular algorithms of brain extraction with scores of 0.913, 0.954 and 0.957 for the Jaccard index, Dice coefficient and sensitivity, respectively. The second part has been evaluated using BRATS database; this methodology has achieved an accuracy of 69% of true detection, and a false detection is around 22% of healthy cases detected as tumour cases and a false detection is around 9% of tumour cases detected as healthy cases. So, the tumour segmentation performed 0.67 Jaccard index and 0.69 Dice coefficient. Our methodology is found to be a fast, effective, accurate and fully automatic one without the need to any human interaction and prior knowledge for training phases as supervised methodologies in clinical applications.

Nonlinear orbital uncertainty propagation with differential algebra and Gaussian mixture model

Nonlinear uncertainty propagation is of critical importance in many application fields of astrodynamics. In this article, a framework combining the differential algebra technique and the Gaussian mixture model method is presented to accurately propagate the state uncertainty of a nonlinear system. A high-order Taylor expansion of the final state with respect to the initial deviations is firstly computed with the differential algebra technique. Then the initial uncertainty is split to a Gaussian mixture model. With the high-order state transition polynomial, each Gaussian mixture element is propagated to the final time, forming the final Gaussian mixture model. Through this framework, the final Gaussian mixture model can include the effects of high-order terms during propagation and capture the non-Gaussianity of the uncertainty, which enables a precise propagation of probability density. Moreover, the manual derivation and integration of the high-order variational equations is avoided, which makes the method versatile. The method can handle both the application of nonlinear analytical maps on any domain of interest and the propagation of initial uncertainties through the numerical integration of ordinary differential equation. The performance of the resulting tool is assessed on some typical orbital dynamic models, including the analytical Keplerian motion, the numerical J2 perturbed motion, and a nonlinear relative motion.

A data mining and item response mixture modeling method to retrospectively measure Diagnostic and Statistical Manual of Mental Disorders-5 attention deficit hyperactivity disorder in the 1970 British Cohort Study.

To facilitate future outcome studies, we aimed to develop a robust and replicable method for estimating a categorical and dimensional measure of Diagnostic and Statistical Manual of Mental Disorders-5 (DSM-5) attention deficit hyperactivity disorder (ADHD) in the 1970 British Cohort Study (BCS70). Following a data mining framework, we mapped DSM-5 ADHD symptoms to age 10 BCS70 data (N=11,426) and derived a 16-item scale (α=0.85). Mapping was validated by an expert panel. A categorical subgroup was derived (n=594, 5.2%), and a zero-inflated item response theory (IRT) mixture model fitted to estimate a dimensional measure. Subgroup composition was comparable with other ADHD samples. Relative risk ratios (ADHD/not ADHD) included boys=1.38, unemployed fathers=2.07, below average reading=2.58, and depressed parent=3.73. Our estimated measures correlated with two derived reference scales: Strengths and Difficulties Questionnaire hyperactivity (r=0.74) and a Rutter/Conners-based scale (r=0.81), supporting construct validity. IRT model items (symptoms) had moderate to high discrimination (0.90-2.81) and provided maximum information at average to moderate theta levels of ADHD (0.5-1.75). We extended previous work to identify ADHD in BCS70, derived scales from existing data, modeled ADHD items with IRT, and adjusted for a zero-inflated distribution. Psychometric properties were promising, and this work will enable future studies of causal mechanisms in ADHD.

Laplacian mixture modeling for network analysis and unsupervised learning on graphs

Laplacian mixture models identify overlapping regions of influence in unlabeled graph and network data in a scalable and computationally efficient way, yielding useful low-dimensional representations. By combining Laplacian eigenspace and finite mixture modeling methods, they provide probabilistic or fuzzy dimensionality reductions or domain decompositions for a variety of input data types, including mixture distributions, feature vectors, and graphs or networks. Provable optimal recovery using the algorithm is analytically shown for a nontrivial class of cluster graphs. Heuristic approximations for scalable high-performance implementations are described and empirically tested. Connections to PageRank and community detection in network analysis demonstrate the wide applicability of this approach. The origins of fuzzy spectral methods, beginning with generalized heat or diffusion equations in physics, are reviewed and summarized. Comparisons to other dimensionality reduction and clustering methods for challenging unsupervised machine learning problems are also discussed.

A Bayesian Approach to Assess Prenatal Exposures to Metal Mixtures and Associations with Latent Patterns of Neurodevelopmental Trajectories

Due to the rise of exposomic research, there is increasing interest in complex mixtures which reflect real life exposure scenarios. Currently, there is limited literature on how exposure to metal mixtures affect neurodevelopmental trajectories. Furthermore, the few existing studies typically do not account for heterogeneity of response to exposure among children, and instead assume that exposure to metal mixtures impacts each child’s health to the same degree. In reality, there may be latent subgroups of children with distinct neurodevelopment trajectories, and the impacts of exposure to metal mixtures may differ across these subgroups.We investigate these questions by developing a two-stage statistical model combining growth mixture modeling and Bayesian kernel-learning methods. This novel statistical approach first identifies latent subgroups with growth mixture modeling using time-varying neurodevelopment scores. In a second step, we use a new statistical model called Bayesian varying coefficient kernel machine regression (BVCKMR), to separately analyze each latent subgroup to determine how metal mixture exposures at a single exposure window affect neurodevelopmental trajectories, while adjusting for covariates. BVCKMR can handle dozens of metal mixture components, allowing us to visualize interactions between up to four metal mixture components, while existing studies typically only focus on pairwise interactions.We illustrate this approach by applying it to the Programming Research in Obesity, Growth, Environment and Social Stressors (PROGRESS) prospective cohort study based in Mexico City. We will present the association between prenatal metal mixtures and child neurodevelopment, as assessed using the Bayley Scales of Infant and Toddler Development. We demonstrate that the relative importance and interaction patterns between metal mixture components differs by latent class, and show that latent classes may be differentially vulnerable to metal mixtures.

Assessing the Influence of in Utero Co-Pollutant Exposures on Transcriptomic Networks and Birth Outcomes

Multiple studies have linked intrauterine exposures to birth outcomes. However, inconsistencies in the literature persist. While heterogeneity in study conditions likely play a role, the discrepancy in reported findings may also reflect a focus on assessing the isolated effect of individual pollutants. Such methods fail to account for the presence of co-pollutants that may alter the activity of the specific exposure under consideration.The molecular pathways through which environmental exposures exert their effects have not been clearly delineated. However, it is suggested that in utero toxicant exposures may induce aberrations in processes mediated by the placenta, the organ overseeing appropriate fetal development. Alterations in the expression and methylation of several placental loci, including genes involved in nutrient transport, endocrine signaling and imprinting, have been linked to birth outcomes. However, similar to studies linking pollutants to birth outcomes, molecular biomarker studies largely focus on univariate associations between individual genes and health effects.High-throughput assessments of environmental exposures and molecular markers are becoming increasingly available to better capture the relationship between toxicant exposures, affected molecular processes and health effects within an environmental epidemiology context. To take advantage of this rich resource, however, methodologies need to be developed that not only account for the complexity but also facilitate integration across datasets. Here, we motivate the application of novel statistical and bioinformatics-based approaches, including environmental mixture modeling methods and gene coexpression networks, to gain insight into the inter-relationship among intrauterine trace metal exposure, placental gene activity and fetal growth.

Memberships of the Open Cluster NGC 6405 Based on a Combined Method: Gaussian Mixture Model and Random Forest

Abstract This paper presents a combined method of Gaussian mixture model and random forest to compute membership probabilities of stars by using large, high-dimensional data sets. A significant advantage of this method is that it allows us to easily identify likely cluster members in large data sets starting from small training samples. As a benchmark, we select 40318 stars in the field of the open cluster NGC 6405 from the Gaia Data Release 2 (Gaia-DR2) by means of all five astrometric (positions, proper motions, and parallax) and photometric parameters. We use this combined method to determine likely cluster members in an eleven-dimensional parameter space. A total number of 518 high-probability (≥0.6) memberships are obtained, and the mean parallax and proper motion of the cluster are determined to be 2.171 ± 0.005 mas (461 ± 1 pc) and ( ) = (−1.357 ± 0.023,−5.823 ± 0.020) mas yr−1, respectively. In addition, we quantitatively evaluate the relative importance of the parameters for membership determination and find that colors and magnitudes cannot be ignored in membership determination when using the RF method. Our results show that this combined method exhibits good performance in handling arbitrary high-dimensional and large data sets, such as Gaia-DR2, and it can also be used to investigate other open clusters.