Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Likelihood Functions”

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.

At least 73 records · Page 4

ANALYSIS OF METHODS FOR DETECTING THE PROXIMITY EFFECT IN QUASAR SPECTRA

Using numerical simulations of structure formation, we investigate several methods for determining the strength of the proximity effect in the H I Ly{alpha} forest. We analyze three high-resolution ({approx}10 kpc) redshift snapshots (z-bar=4,3, and 2.25) of a Hydro-Particle-Mesh simulation to obtain realistic absorption spectra of the H I Ly{alpha} forest. We model the proximity effect along the simulated sight lines with a simple analytical prescription based on the assumed quasar luminosity and the intensity of the cosmic UV background (UVB). We begin our analysis investigating the intrinsic biases thought to arise in the widely adopted standard technique of combining multiple lines of sight when searching for the proximity effect. We confirm the existence of these biases, albeit smaller than previously predicted with simple Monte Carlo simulations. We then concentrate on the analysis of the proximity effect along individual lines of sight. After determining its strength with a fiducial value of the UVB intensity, we construct the proximity effect strength distribution (PESD). We confirm that the PESD inferred from the simple averaging technique accurately recovers the input strength of the proximity effect at all redshifts. Moreover, the PESD closely follows the behaviors found in observed samples of quasar spectra. However, the PESD obtained from our new simulated sight lines presents some differences to that of simple Monte Carlo simulations. At all redshifts, we find a smaller dispersion of the strength parameters, the source of the corresponding smaller biases found when combining multiple lines of sight. After developing three new theoretical methods for recovering the strength of the proximity effect on individual lines of sight, we compare their accuracy to the PESD from the simple averaging technique. All our new approaches are based on the maximization of the likelihood function, albeit invoking some modifications. The new techniques presented here, in spite of their complexity, fail to recover the input proximity effect in an unbiased way, presumably due to some (unknown) higher order correlations in the spectrum. Thus, employing complex three-dimensional simulations, we provide strong evidence in favor of the PESD obtained from the simple averaging technique, as a method of estimating the UVB intensity, free of any intrinsic biases.

79 ASTRONOMY AND ASTROPHYSICS↗

Optimal bispectrum constraints on single-field models of inflation

We use WMAP 9-year bispectrum data to constrain the free parameters of an 'effective field theory' describing fluctuations in single-field inflation. The Lagrangian of the theory contains a finite number of operators associated with unknown mass scales. Each operator produces a fixed bispectrum shape, which we decompose into partial waves in order to construct a likelihood function. Based on this likelihood we are able to constrain four linearly independent combinations of the mass scales. As an example of our framework we specialize our results to the case of 'Dirac-Born-Infeld' and 'ghost' inflation and obtain the posterior probability for each model, which in Bayesian schemes is a useful tool for model comparison. Our results suggest that DBI-like models with two or more free parameters are disfavoured by the data by comparison with single-parameter models in the same class.

79 ASTRONOMY AND ASTROPHYSICS↗

CLUSTERING OF SLOAN DIGITAL SKY SURVEY III PHOTOMETRIC LUMINOUS GALAXIES: THE MEASUREMENT, SYSTEMATICS, AND COSMOLOGICAL IMPLICATIONS

The Sloan Digital Sky Survey (SDSS) surveyed 14,555 deg{sup 2}, and delivered over a trillion pixels of imaging data. We present a study of galaxy clustering using 900,000 luminous galaxies with photometric redshifts, spanning between z = 0.45 and z = 0.65, constructed from the SDSS using methods described in Ross et al. This data set spans 11,000 deg{sup 2} and probes a volume of 3 h {sup -3} Gpc{sup 3}, making it the largest volume ever used for galaxy clustering measurements. We describe in detail the construction of the survey window function and various systematics affecting our measurement. With such a large volume, high-precision cosmological constraints can be obtained given careful control and understanding of the observational systematics. We present a novel treatment of the observational systematics and its applications to the clustering signals from the data set. In this paper, we measure the angular clustering using an optimal quadratic estimator at four redshift slices with an accuracy of {approx}15%, with a bin size of {delta}{sub l} = 10 on scales of the baryon acoustic oscillations (BAOs; at l {approx} 40-400). We also apply corrections to the power spectra due to systematics and derive cosmological constraints using the full shape of the power spectra. For a flat {Lambda}CDM model, when combined with cosmic microwave background Wilkinson Microwave Anisotropy Probe 7 (WMAP7) and H{sub 0} constraints from using 600 Cepheids observed by Wide Field Camera 3 (WFC3; HST), we find {Omega}{sub {Lambda}} = 0.73 {+-} 0.019 and H{sub 0} to be 70.5 {+-} 1.6 s{sup -1} Mpc{sup -1} km. For an open {Lambda}CDM model, when combined with WMAP7 + HST, we find {Omega}{sub K} = 0.0035 {+-} 0.0054, improved over WMAP7+HST alone by 40%. For a wCDM model, when combined with WMAP7+HST+SN, we find w = -1.071 {+-} 0.078, and H{sub 0} to be 71.3 {+-} 1.7 s{sup -1} Mpc{sup -1} km, which is competitive with the latest large-scale structure constraints from large spectroscopic surveys such as the SDSS Data Release 7 (DR7) and WiggleZ. We also find that systematic-corrected power spectra give consistent constraints on cosmological models when compared with pre-systematic correction power spectra in the angular scales of interest. The SDSS-III Data Release 8 (SDSS-III DR8) Angular Clustering Data allow a wide range of investigations into the cosmological model, cosmic expansion (via BAO), Gaussianity of initial conditions, and neutrino masses. Here, we refer to our companion papers for further investigations using the clustering data. Our calculation of the survey selection function, systematics maps, and likelihood function for the COSMOMC package will be released at http://portal.nersc.gov/project/boss/galaxy/photoz/.

79 ASTRONOMY AND ASTROPHYSICS↗

Pattern recognition principles

The present work gives an account of basic principles and available techniques for the analysis and design of pattern processing and recognition systems. Areas covered include decision functions, pattern classification by distance functions, pattern classification by likelihood functions, the perceptron and the potential function approaches to trainable pattern classifiers, statistical approach to trainable classifiers, pattern preprocessing and feature selection, and syntactic pattern recognition.

Tou, J. T.↗

Modeling of solar oscillation power spectra

To produce accurate estimates of the line-profile parameters of a model used to represent the spectral features in a solar oscillation power spectrum, it is necessary to (1) select the appropriate probability density function when deriving the maximum-likelihood function to be employed for the parameter estimation and (2) allow for the redistribution of spectral power caused by gaps in the data string. This paper describes a maximum-likelihood method for estimating the model parameters (based on the observed power spectrum statistics) that accounts for redistribution of spectral power caused by gaps in the data string, by convolving the model with the power spectrum of the observed window function. The accuracy and reliability of the method were tested using both artificial and authentic solar oscillation power spectrum data. A comparison of this method with various least-squares techniques is also presented.

Anderson, Edwin R.↗

Image Moment-Based Extended Object Tracking for Complex Motions

A novel image moment-based model for shape estimation and tracking of an extended target moving with a complex trajectory is presented. The proposed extended object tracking algorithm is based on multiple noisy measurement points sampled from the target at each time step. The shape of the object, approximated by an ellipse, is estimated using a combination of image moments. Dynamic models of image moments for constant velocity and coordinated turn motions are mathematically derived. An unscented Kalman filter - interacting multiple model (UKF-IMM) method is used to track the object and estimate its shape. A likelihood function based on average log-likelihood is derived for the IMM filter. Simulation results of the proposed UKF-IMM algorithm with the image momentbased models are presented that show the estimation of the shape of the object moving in a complex trajectory. The intersection over union (IoU), and the root mean square errors (RMSEs) of the position and velocity of the centroid of the ellipse are used as metrics. The comparison results of the proposed algorithm with a benchmark algorithm from literature based on the IoU and RMSE metrics are presented.

Extended object tracking↗

Statistically Rigorous Uncertainty Quantification for Physical Parameter Model Calibration with Functional Output

In experiments conducted on the Z-machine at Sandia National Laboratories, dynamic material properties cannot be analyzed using traditional analytic methods, necessitating solving an inverse problem. Bayesian model calibration is a statistical framework for solving an inverse problem to estimate parameters input into a computational model in the presence of multiple uncertainties. Disentangling input parameter uncertainty and model misspecification is often poorly identified problem. When using computational models for physical parameter estimation, the issue of parameter identifiability must be carefully considered to obtain accurate and precise estimates of physical parameters. Additionally, in dynamic material properties applications, the experimental output is a function, velocity over time. While we can sample an arbitrarily large number of points from the measured velocity, these curves only contain a finite amount of information about the calibration parameters. In this report, we propose modifications to the Bayesian model calibration framework to simplify and improve the estimation of physical parameters with functional outputs. Specifically, we propose scaling the likelihood function by an effective sample size rather than modeling the discrepancy function; and modularizing input nuisance parameters with weakly identified parameters. We evaluate the performance of these proposed methods using a statistical simulation study and then apply these methods to estimate parameters of the tantalum equation of state. We conclude that these proposed methods can provide simple, fast, and statistically valid alternatives to the full Bayesian model calibration procedure; and that these methods can be used to estimate parameters of the equation of state for tantalum.

97 MATHEMATICS AND COMPUTING↗

A Convex Optimization Approach to Improving Suboptimal Hyperparameters of Sliced Normal Distributions

Sliced Normal (SN) distributions are a generalization of Gaussian distributions where the quadratic argument of the exponential is replaced with a sum of squares polynomial. SNs may be used to represent the distribution of a diverse set of random variables including multi-modal, non-symmetric, and skewed distributions. Unfortunately, the likelihood function of a SN includes a normalization constant and the inclusion of this normalization constant makes the likelihood a non-convex function of the hyperparameters which define the SN. In previous work, suboptimal fitting of the hyperparameters was performed by transforming the given data into a higher dimensional monomial basis and selecting the optimal hyperparameters of a Gaussian fit in this space. However, this approach did not account for the effect of lifting on the normalization constant. Indeed, it was observed that as the number of monomials is increased the likelihood of the Sliced Normal can decrease. In this paper, we increase the likelihood of Sliced Normals found using the previous method by developing a convex formulation which scales the covariance matrix of the Gaussian fit such that the likelihood of the Sliced Normal is maximized. The result is significant improvements of the log likelihood of fitted SN distributions, including a significant increase, especially for problems with 500+ monomials.

Convex optimization approach to improving suboptim↗

Stochastic control system parameter identifiability

The parameter identification problem of general discrete time, nonlinear, multiple input/multiple output dynamic systems with Gaussian white distributed measurement errors is considered. The knowledge of the system parameterization was assumed to be known. Concepts of local parameter identifiability and local constrained maximum likelihood parameter identifiability were established. A set of sufficient conditions for the existence of a region of parameter identifiability was derived. A computation procedure employing interval arithmetic was provided for finding the regions of parameter identifiability. If the vector of the true parameters is locally constrained maximum likelihood (CML) identifiable, then with probability one, the vector of true parameters is a unique maximal point of the maximum likelihood function in the region of parameter identifiability and the constrained maximum likelihood estimation sequence will converge to the vector of true parameters.

Lee, C. H.↗

Testing the Responses of Three PSF Fitting Methods

We present an analysis of prototype code designed to fit point-spread functions to point-like astronomical objects appearing in images taken by DECam, or the Dark Energy Camera. We test three methods, each of which yields a maximum likelihood estimate for a different likelihood function. We call these methods Poisson likelihood, χ 2 with background error, and χ 2 with data error, and describe their differences mathematically. Each is used to construct a fit involving up to four parameters characterizing the raw pixel data. We measure biases in the signal and background levels under a variety of conditions. We test the variance and error estimate of the fits as a function of signal strength, outline areas of sensitivity, and demonstrate the effects of incomplete or inaccurate information on the Poisson fit. We conclude that the Poisson likelihood fitting method is best, as it is successful in all circumstances under which the χ 2 series are, and displays other favorable attributes where the others do not.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Measurement of the ratio $B(t \to W b)/B(t \to W q)$ in $t\bar{t}$ dilepton channel at CDF

My analysis is based on the number of b-jets found in t¯t events using the dilepton sample with at least 2 jets in the final state. The charged leptons could be either electrons or muons. Tau leptons are not included. We use SecVtx algorithm, based on the reconstruction of a secondary vertex in the event, in order to identify a jet coming from b-quark fragmentation (b-tagging). Due to the high purity of the t¯t signal in dilepton events it is possible to perform a kinematic measurement of the t¯t cross section. Our strategy is to use this result to make prediction on the number of t¯t events. We divide our sample in subsets according to dilepton type (combination of the lepton type), number of jets in the final states and events with zero, one or two tags. The comparison between events and the prediction, given by the sum of the expected t¯t estimate and the background yield, in each subsample is made using a Likelihood function. Our measured value for R is the one which maximizes the Likelihood, i.e. gives the best match between our expectation and the observed data. We measure: p¯p!t¯t = 7.05±0.53stat±0.42lumi , R= 0.86±0.06 (stat+syst) and, in the hypothesis of CKM matrix unitarity with three quark generations, | Vtb | = 0.93 ± 0.03. Our analysis on the p¯p!t¯t was performed independently of the official dilepton analisys on the t¯t production cross section. So it represents also a valuable crosscheck for the official analysis. In chapter 1, a brief introduction to the theoretical framework is given. The standard model of elementary particles and the Quantum Cromodynamic theories are introduced. Then the top quark is presented, with a short descpription of its properties, as its mass, its production mode and its cross section. Some previous results on R are listed as well. Later we present the experiment that collected our data, both the collider (chapter 2) and the detector (CDF)(chapter 3). In chapter 4 we describe the physics object reconstruction, so how we collect the detector signals and translate those into physical particles traversing our detector. The event selection is described in chapter 5, where we report the complete list of the selection requirements and we estimate our sample composition. In chapters 6 and 7 we report our results for the t¯t production cross section and R. An indirect measurement for |Vtb| is given as well.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Estimation of correlation functions by stochastic approximation.

Consideration of the autocorrelation function of a zero-mean stationary random process. The techniques are applicable to processes with nonzero mean provided the mean is estimated first and subtracted. Two recursive techniques are proposed, both of which are based on the method of stochastic approximation and assume a functional form for the correlation function that depends on a number of parameters that are recursively estimated from successive records. One technique uses a standard point estimator of the correlation function to provide estimates of the parameters that minimize the mean-square error between the point estimates and the parametric function. The other technique provides estimates of the parameters that maximize a likelihood function relating the parameters of the function to the random process. Examples are presented.

Habibi, A.↗

Cluster Sampling Filters for Non-Gaussian Data Assimilation

This paper presents a fully non-Gaussian filter for sequential data assimilation. The filter is named the “cluster sampling filter”, and works by directly sampling the posterior distribution following a Markov Chain Monte-Carlo (MCMC) approach, while the prior distribution is approximated using a Gaussian Mixture Model (GMM). Specifically, a clustering step is introduced after the forecast phase of the filter, and the prior density function is estimated by fitting a GMM to the prior ensemble. Using the data likelihood function, the posterior density is then formulated as a mixture density, and is sampled following an MCMC approach. Four versions of the proposed filter, namely C ℓ MCMC , C ℓ HMC , MC- C ℓ HMC , and MC- C ℓ HMC are presented. C ℓ MCMC uses a Gaussian proposal density to sample the posterior, and C ℓ HMC is an extension to the Hamiltonian Monte-Carlo (HMC) sampling filter. MC- C ℓ MCMC and MC- C ℓ HMC are multi-chain versions of the cluster sampling filters C ℓ MCMC and C ℓ HMC respectively. The multi-chain versions are proposed to guarantee that samples are taken from the vicinities of all probability modes of the formulated posterior. The new methodologies are tested using a simple one-dimensional example, and a quasi-geostrophic (QG) model with double-gyre wind forcing and bi-harmonic friction. Here, numerical results demonstrate the usefulness of using GMMs to relax the Gaussian prior assumption especially in the HMC filtering paradigm.

97 MATHEMATICS AND COMPUTING↗

Statistical analysis of failure data on controllers and SSME turbine blade failures

The expressions for the maximum likelihood functions are given when the failure data are censored at a given point or at multiple points, or when the data come in groups. Different models applicable to failure data are presented with their characteristics. A graphical method of distinguishing different models by using cumulative hazard fucnction is discussed. For the failure data on controllers the model is determined by cumulative hazard function and chi-square goodness of fit. Using the Weibull Model the maximum likelihood estimators of the shape parameter and the failure rate parameter are obtained. The confidence intervals, meantime between failures, and B1 are determined. Similarly, for the data on Space Shuttle Main Engine (SSME) blade failures the maximum likelihood estimators are obtained for the Weibull parameters. The variances, confidence intervals, meantime between failures, and reliability are determined. The analysis is performed under assumption of grouped data as well as randomly placed data.

Patil, S. A.↗

Spectral Components Analysis of Diffuse Emission Processes

We develop a novel method to separate the components of a diffuse emission process based on an association with the energy spectra. Most of the existing methods use some information about the spatial distribution of components, e.g., closeness to an external template, independence of components etc., in order to separate them. In this paper we propose a method where one puts conditions on the spectra only. The advantages of our method are: 1) it is internal: the maps of the components are constructed as combinations of data in different energy bins, 2) the components may be correlated among each other, 3) the method is semi-blind: in many cases, it is sufficient to assume a functional form of the spectra and determine the parameters from a maximization of a likelihood function. As an example, we derive the CMB map and the foreground maps for seven yeas of WMAP data. In an Appendix, we present a generalization of the method, where one can also add a number of external templates.

79 ASTRONOMY AND ASTROPHYSICS↗

Free energy reconstruction from steered dynamics without post-processing

Various methods achieving importance sampling in ensembles of nonequilibrium trajectories enable one to estimate free energy differences and, by maximum-likelihood post-processing, to reconstruct free energy landscapes. Here, based on Bayes theorem, we propose a more direct method in which a posterior likelihood function is used both to construct the steered dynamics and to infer the contribution to equilibrium of all the sampled states. The method is implemented with two steering schedules. First, using non-autonomous steering, we calculate the migration barrier of the vacancy in Fe-{alpha}. Second, using an autonomous scheduling related to metadynamics and equivalent to temperature-accelerated molecular dynamics, we accurately reconstruct the two-dimensional free energy landscape of the 38-atom Lennard-Jones cluster as a function of an orientational bond-order parameter and energy, down to the solid-solid structural transition temperature of the cluster and without maximum-likelihood post-processing.

97 MATHEMATICS AND COMPUTING↗

Multiple object tracking with non-unique data-to-object association via generalized hypothesis testing

A generalized hypothesis testing approach is applied to the problem of tracking several objects where several different associations of data with objects are possible. Such problems occur, for instance, when attempting to distinctly track several aircraft maneuvering near each other or when tracking ships at sea. Conceptually, the problem is solved by first, associating data with objects in a statistically reasonable fashion and then, tracking with a bank of Kalman filters. The objects are assumed to have motion characterized by a fixed but unknown deterministic portion plus a random process portion modeled by a shaping filter. For example, the object might be assumed to have a mean straight line path about which it maneuvers in a random manner. Several hypothesized associations of data with objects are possible because of ambiguity as to which object the data comes from, false alarm/detection errors, and possible uncertainty in the number of objects being tracked. The statistical likelihood function is computed for each possible hypothesized association of data with objects. Then the generalized likelihood is computed by maximizing the likelihood over parameters that define the deterministic motion of the object.

Porter, D. W.↗

Directional Statistics of Preferential Orientations of Two Shapes in Their Aggregate and Its Application to Nanoparticle Aggregation

Nanoscientists have long conjectured that adjacent nanoparticles aggregate with one another in certain preferential directions during chemical syntheses of nanoparticles, which is so called the oriented attachment. For the study of the oriented attachment phenomenon, the microscopy and nanoscience community have used dynamic electron microscopy for direct observations of nanoparticle aggregation and have been so far relying on the manual and qualitative analysis of the observations. In this paper, we present a statistical approach for studying the oriented attachment phenomenon with multiple aggregation examples observed by dynamic microscope data. We abstract a nanoparticle aggregation by an event of two primary geometric objects merging into a secondary geometric object. We define a metric space for geometrical bodies, and the endowed metric in the space allows us to define the likelihood function of the orientation angles of two primary objects within the corresponding secondary object. We maximize the likelihood to estimate the orientation angles. We fit a mixture of bivariate von Mises distributions to the estimated orientation angles for the density estimation. By analyzing the density estimate, we are able to demonstrate that nanoparticles of certain geometries have indeed preferential orientations in their aggregates.

Mixture of bivariate von Mises distributions, Bi-a↗