SEARCH · Engineering Papers
Results for “Finite Mixture Models”
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Predicting Damages to Remainder Parcels in Right-of-Way Acquisitions for Expanding Transportation Infrastructure: Using a Truncated Finite-Mixture Model
Right-of-way acquisition is a critical component of transportation infrastructure development. Transportation infrastructure projects cannot proceed without proper right-of-way acquisition or may face significant delays. State Departments of Transportation frequently acquire parcels of land for roadway expansion projects. A majority of these acquisitions can be partial takings, referring to a portion of a parcel that is acquired. The remainder of the property usually suffers economic changes due to the partial acquisition, which can be calculated as damage percentages. The damage percentage represents the extent to which the remaining land or property value has been diminished due to the acquisition. It reflects the remaining property value percentage that may have been lost or compromised due to the acquisition. Here, this study aims to provide a robust model to estimate damage percentages to the remainder parcels that may help state Departments of Transportation appraisers make early predictions about the damages in cases involving partial takings. The research uses 509 appraisal reports from the Tennessee Department of Transportation to identify the key parcel attributes that influence the percentage of damages. Three regression models are developed: a linear regression model, a finite-mixture model (FMM), and a truncated FMM with two latent classes. The modeling results show that the truncated FMM with two classes outperforms the other models. To validate the models, actual sales data is collected and analyzed for 59 properties, and the results suggest that the model predictions are fairly accurate. A predictive tool is developed based on the models to help appraisers anticipate right-of-way damages under different scenarios and can provide early predictions about the damages.
Disaggregating power consumption of commercial buildings based on the finite mixture model
Not Available
MixSim : An R Package for Simulating Data to Study Performance of Clustering Algorithms
The R package MixSim is a new tool that allows simulating mixtures of Gaussian distributions with different levels of overlap between mixture components. Pairwise overlap, defined as a sum of two misclassification probabilities, measures the degree of interaction between components and can be readily employed to control the clustering complexity of datasets simulated from mixtures. These datasets can then be used for systematic performance investigation of clustering and finite mixture modeling algorithms. Among other capabilities of MixSim, there are computing the exact overlap for Gaussian mixtures, simulating Gaussian and non-Gaussian data, simulating outliers and noise variables, calculating various measures of agreement between two partitionings, and constructing parallel distribution plots for the graphical display of finite mixture models. All features of the package are illustrated in great detail. The utility of the package is highlighted through a small comparison study of several popular clustering algorithms.
Model-Based Clustering of Regression Time Series Data via APECM -- An AECM Algorithm Sung to an Even Faster Beat
We propose a model-based approach for clustering time series regression data in an unsupervised machine learning framework to identify groups under the assumption that each mixture component follows a Gaussian autoregressive regression model of order p. Given the number of groups, the traditional maximum likelihood approach of estimating the parameters using the expectation-maximization (EM) algorithm can be employed, although it is computationally demanding. The somewhat fast tune to the EM folk song provided by the Alternating Expectation Conditional Maximization (AECM) algorithm can alleviate the problem to some extent. In this article, we develop an alternative partial expectation conditional maximization algorithm (APECM) that uses an additional data augmentation storage step to efficiently implement AECM for finite mixture models. Results on our simulation experiments show improved performance in both fewer numbers of iterations and computation time. The methodology is applied to the problem of clustering mutual funds data on the basis of their average annual per cent returns and in the presence of economic indicators.
Inference of the phase-to-mechanical property link via coupled X-ray spectrometry and indentation analysis: Application to cement-based materials
A novel approach for the chemo-mechanical characterization of cement-based materials is presented, which combines the classical grid indentation technique with elemental mapping by scanning electron microscopy-energy dispersive X-ray spectrometry (SEM-EDS). It is illustrated through application to an oil-well cement system with siliceous filler. The characteristic X-rays of major elements (silicon, calcium and aluminum) are measured over the indentation region and mapped back on the indentation points. Measured intensities together with indentation hardness and modulus are considered in a clustering analysis within the framework of Finite Mixture Models with Gaussian component density function. The method is able to successfully isolate the calcium-silica-hydrate gel at the indentation scale from its mixtures with other products of cement hydration and anhydrous phases; thus providing a convenient means to link mechanical response to the calcium-to-silicon ratio quantified independently via X-ray wavelength dispersive spectroscopy. A discussion of uncertainty quantification of the estimated chemo-mechanical properties and phase volume fractions, as well as the effect of chemical observables on phase assessment is also included.
The spatial structure of young stellar clusters. I. Subclusters
The clusters of young stars in massive star-forming regions show a wide range of sizes, morphologies, and numbers of stars. Their highly subclustered structures are revealed by the MYStIX project's sample of 31,754 young stars in nearby sites of star formation (regions at distances <3.6 kpc that contain at least one O-type star.) In 17 of the regions surveyed by MYStIX, we identify subclusters of young stars using finite mixture models—collections of isothermal ellipsoids that model individual subclusters. Maximum likelihood estimation is used to estimate the model parameters, and the Akaike Information Criterion is used to determine the number of subclusters. This procedure often successfully finds famous subclusters, such as the BN/KL complex behind the Orion Nebula Cluster and the KW-object complex in M 17. A catalog of 142 subclusters is presented, with 1-20 subclusters per region. The subcluster core radius distribution for this sample is peaked at 0.17 pc with a standard deviation of 0.43 dex, and subcluster core radius is negatively correlated with gas/dust absorption of the stars—a possible age effect. Based on the morphological arrangements of subclusters, we identify four classes of spatial structure: long chains of subclusters, clumpy structures, isolated clusters with a core-halo structure, and isolated clusters well fit by a single isothermal ellipsoid.
Expectation-maximization algorithms for learning a finite mixture of univariate survival time distributions from partially specified class values
Heterogeneity exists on a data set when samples from di erent classes are merged into the data set. Finite mixture models can be used to represent a survival time distribution on heterogeneous patient group by the proportions of each class and by the survival time distribution within each class as well. The heterogeneous data set cannot be explicitly decomposed to homogeneous subgroups unless all the samples are precisely labeled by their origin classes; such impossibility of decomposition is a barrier to overcome for estimating nite mixture models. The expectation-maximization (EM) algorithm has been used to obtain maximum likelihood estimates of nite mixture models by soft-decomposition of heterogeneous samples without labels for a subset or the entire set of data. In medical surveillance databases we can find partially labeled data, that is, while not completely unlabeled there is only imprecise information about class values. In this study we propose new EM algorithms that take advantages of using such partial labels, and thus incorporate more information than traditional EM algorithms. We particularly propose four variants of the EM algorithm named EM-OCML, EM-PCML, EM-HCML and EM-CPCML, each of which assumes a specific mechanism of missing class values. We conducted a simulation study on exponential survival trees with five classes and showed that the advantages of incorporating substantial amount of partially labeled data can be highly signi cant. We also showed model selection based on AIC values fairly works to select the best proposed algorithm on each specific data set. A case study on a real-world data set of gastric cancer provided by Surveillance, Epidemiology and End Results (SEER) program showed a superiority of EM-CPCML to not only the other proposed EM algorithms but also conventional supervised, unsupervised and semi-supervised learning algorithms.
A Learning Scheme For Recognizing Sub-classes From Model Trained On Aggregate Classes
In many practical situations it is not feasible to collect labeled samples for all available classes in a domain. Especially in supervised classification of remotely sensed images it is impossible to collect ground truth information over large geographic regions for all thematic classes. As a result often analysts collect labels for aggregate classes. In this paper we present a novel learning scheme that automatically learns sub-classes from the user given aggregate classes. We model each aggregate class as finite Gaussian mixture instead of classical assumption of unimodal Gaussian per class. The number of components in each finite Gaussian mixture are automatically estimated. Experimental results on real remotely sensed image classification showed not only improved accuracy in aggregate class classification but the proposed method also recognized sub-classes.
Monte Carlo goodness-of-fit tests for degree corrected and related stochastic blockmodels
Abstract We construct Bayesian and frequentist finite-sample goodness-of-fit tests for three different variants of the stochastic blockmodel for network data. Since all of the stochastic blockmodel variants are log-linear in form when block assignments are known, the tests for the latent block model versions combine a block membership estimator with the algebraic statistics machinery for testing goodness-of-fit in log-linear models. We describe Markov bases and marginal polytopes of the variants of the stochastic blockmodel and discuss how both facilitate the development of goodness-of-fit tests and understanding of model behaviour. The general testing methodology developed here extends to any finite mixture of log-linear models on discrete data, and as such is the first application of the algebraic statistics machinery for latent-variable models.
On the calculation of equilibrium thermodynamic properties and the establishment of statistical-thermodynamically-consistent finite bound-state partition functions in nonideal multicomponent plasma mixtures within the chemical model
The problem of the calculation of equilibrium thermodynamic properties and the establishment of statistical-thermodynamically consistent finite bound-state partition functions in nonideal multicomponent plasma systems is revised within the chemical picture. The present exploration accompanied by the introduction of a generalized consistent formulation, in terms of the solution of the inverse problem, clears ambiguities and gives a better understanding of the problem on top of pointing out weaknesses and inaccuracies/inconsistencies buried in widely used models in literature.
Sub-class Recognition from Aggregate Class Labels: Preliminary Results
In many practical situations it is not feasible to collect labeled samples for all available classes in a domain. Especially in supervised classification of remotely sensed images it is impossible to collect ground truth information over large geographic regions for all thematic classes. As a result often analysts collect labels for aggregate classes. In this paper we present a novel learning scheme that automatically learns sub-classes from the user given aggregate classes. We model each aggregate class as finite Gaussian mixture instead of classical assumption of unimodal Gaussian per class. The number of components in each finite Gaussian mixture are automatically estimated. Experimental results on real remotely sensed image classification showed not only improved accuracy in aggregate class classification but the proposed method also recognized sub-classes.
On a stochastic approach to a code performance estimation
The main goal of an efficient profiling of software is to minimize the runtime overhead under certain constraints and requirements. The traces built by a profiler during the work, affect the performance of the system itself. One of important aspect of an overhead arises from the randomness of variability in the context in which the application is embedded, e.g., due to possible cache misses, etc. Such uncertainty needs to be taken into account in the design phase. In order to overcome these difficulties we propose to investigate this issue through the analysis of the probability distribution of the difference between profiler’s times for the same code. The approximating model is based on the finite normal mixtures within the framework of the method of moving separation of mixtures. We demonstrate some results for the MATLAB profiler using plotting of 3D surfaces by the function surf. The idea can be used for an estimating of a program efficiency.
A robust estimator of mutual information for deep learning interpretability
Abstract We develop the use of mutual information (MI), a well-established metric in information theory, to interpret the inner workings of deep learning (DL) models. To accurately estimate MI from a finite number of samples, we present GMM-MI (pronounced ‘Jimmie’), an algorithm based on Gaussian mixture models that can be applied to both discrete and continuous settings. GMM-MI is computationally efficient, robust to the choice of hyperparameters and provides the uncertainty on the MI estimate due to the finite sample size. We extensively validate GMM-MI on toy data for which the ground truth MI is known, comparing its performance against established MI estimators. We then demonstrate the use of our MI estimator in the context of representation learning, working with synthetic data and physical datasets describing highly non-linear processes. We train DL models to encode high-dimensional data within a meaningful compressed (latent) representation, and use GMM-MI to quantify both the level of disentanglement between the latent variables, and their association with relevant physical quantities, thus unlocking the interpretability of the latent representation. We make GMM-MI publicly available in this GitHub repository.
Detailed finite element method modeling of evaporating multi-component droplets
The evaporation of sessile multi-component droplets is modeled with an axisymmetic finite element method. The model comprises the coupled processes of mixture evaporation, multi-component flow with composition-dependent fluid properties and thermal effects. Based on representative examples of water–glycerol and water–ethanol droplets, regular and chaotic examples of solutal Marangoni flows are discussed. Furthermore, the relevance of the substrate thickness for the evaporative cooling of volatile binary mixture droplets is pointed out. It is shown how the evaporation of the more volatile component can drastically decrease the interface temperature, so that ambient vapor of the less volatile component condenses on the droplet. Finally, results of this model are compared with corresponding results of a lubrication theory model, showing that the application of lubrication theory can cause considerable errors even for moderate contact angles of 40°. - Graphical abstract:.
The research into the propagation law of the shock wave of a gas explosion inside a building
Based on the dissipation rate conservation equations of turbulent kinetic energy in the k‐ ε turbulence model, a complicated three‐dimensional finite element model of a kitchen filled with gas mixture is developed by using the open source field operation and manipulation (OpenFOAM). Two representative kitchens were used to investigate the propagation law of the shock wave of a gas explosion inside a building by considering the key characteristics of the blast shock wave. The influence of some crucial parameters, such as initial conditions and kitchen parameters, on the properties of the blast shock wave is investigated. The basic steps to predict the peak pressure of the blast shock wave are given in consideration of the initial condition and the kitchen whilst the injury effect of the blast shock wave on the humans and animals is evaluated. The research results indicate that the pressure time history and the peak pressure space distribution are greatly influenced by the kitchen design layout. The coupled interaction between the initial temperature and gas volume concentration, especially at the upper and lower explosion limits of the gas, significantly affects the peak pressure. The peak pressure varies significantly with the opening and the buffer; however, it has little relation with the width, length, and height of the kitchen. The proposed method can accurately and effectively predict the peak pressure of the blast shock wave inside buildings. In terms of the peak pressure space distribution of the explosion shock wave, the peak pressure is much higher than the threshold of the killing pressure, which is unsafe for the humans and animals in the building.