Search NASA⌕ Search

SEARCH · Search NASA

Results for “Bayesian Methods”

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 1,351 records · Page 75

TDCOSMO VIII. A key test of systematics in the hierarchical method of time-delay cosmography

The largest source of systematic errors in the time-delay cosmography method likely arises from the lens model mass distribution, where an inaccurate choice of model could in principle bias the value of $H$ 0 . A Bayesian hierarchical framework has been proposed which combines lens systems with kinematic data, constraining the mass profile shape at a population level. The framework has been previously validated using a small sample of lensing galaxies drawn from hydro-simulations. The goal of this work is to expand the validation to a more general set of lenses consistent with observed systems, as well as confirm the capacity of the method to combine two lens populations: one which has time delay information and one which lacks time delays and has systematically different image radii. For this purpose, we generated samples of analytic lens mass distributions made of baryons+dark matter and fit the subsequent mock images with standard power-law models. Corresponding kinematics data were also emulated. The hierarchical framework applied to an ensemble of time-delay lenses allowed us to correct the $H$ 0 bias associated with model choice to find $H$ 0 within 1.5σ of the fiducial value. We then combined this set with a sample of corresponding lens systems which have no time delays and have a source at lower $z$, resulting in a systematically smaller image radius relative to their effective radius. The hierarchical framework has successfully accounted for this effect, recovering a value of $H$ 0 which is both more precise (σ ~ 2%) and more accurate (0.7% median offset) than the time-delay set alone. This result confirms that non-time-delay lenses can nonetheless contribute valuable constraining power to the determination of $H$ 0 via their kinematic constraints, assuming they come from the same global population as the time-delay set.

79 ASTRONOMY AND ASTROPHYSICS↗

Reference air kerma and kerma-area product as estimators of peak skin dose for fluoroscopically guided interventions

Purpose: To determine more accurate regression formulas for estimating peak skin dose (PSD) from reference air kerma (RAK) or kerma-area product (KAP). Methods: After grouping of the data from 21 procedures into 13 clinically similar groups, assessments were made of optimal clustering using the Bayesian information criterion to obtain the optimal linear regressions of (log-transformed) PSD vs RAK, PSD vs KAP, and PSD vs RAK and KAP. Results: Three clusters of clinical groups were optimal in regression of PSD vs RAK, seven clusters of clinical groups were optimal in regression of PSD vs KAP, and six clusters of clinical groups were optimal in regression of PSD vs RAK and KAP. Prediction of PSD using both RAK and KAP is significantly better than prediction of PSD with either RAK or KAP alone. The regression of PSD vs RAK provided better predictions of PSD than the regression of PSD vs KAP. The partial-pooling (clustered) method yields smaller mean squared errors compared with the complete-pooling method.Conclusion: PSD distributions for interventional radiology procedures are log-normal. Estimates of PSD derived from RAK and KAP jointly are most accurate, followed closely by estimates derived from RAK alone. Estimates of PSD derived from KAP alone are the least accurate. Using a stochastic search approach, it is possible to cluster together certain dissimilar types of procedures to minimize the total error sum of squares.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Analytic Neural Network Gaussian Process Enabled Chance-Constrained Voltage Regulation for Active Distribution Systems with PVs, Batteries and EVs

This paper proposes an analytic neural network Gaussian process (NNGP)-based chance-constrained real-time voltage regulation method for active distribution systems with photovoltaics (PVs), batteries, and electric vehicles (EVs). NNGP can utilize historical measurement data to achieve real-time probabilistic node voltage estimation through Bayesian inference. Then, NNGP is fully analytically embedded into the optimal power flow model to perform voltage regulation and adapt to various topological changes. The uncertainties of voltage estimations are easily considered via the chance constraint, and it has been shown that the adoption of this chance constraint can significantly improve the reliability of voltage regulation under various scenarios. The comparison results with other methods, carried out on a real 759-node distribution system located in western Colorado, U.S., show that the proposed method can achieve accurate voltage estimation across different topologies and reliably perform voltage regulation considering PVs, batteries, and EVs.

active distribution systems↗

Integrating System Dynamics and Bayesian Networks with Application to Counter-IED Scenarios

The practice of choosing a single modeling paradigm for predictive analysis can limit the scope and relevance of predictions and their utility to decision-making processes. Considering multiple modeling methods simultaneously may improve this situation, but a better solution provides a framework for directly integrating different, potentially complementary modeling paradigms to enable more comprehensive modeling and predictions, and thus better-informed decisions. The primary challenges of this kind of model integration are to bridge language and conceptual gaps between modeling paradigms, and to determine whether natural and useful linkages can be made in a formal mathematical manner. To address these challenges in the context of two specific modeling paradigms, we explore mathematical and computational options for linking System Dynamics (SD) and Bayesian network (BN) models and incorporating data into the integrated models. We demonstrate that integrated SD/BN models can naturally be described as either state space equations or Dynamic Bayes Nets, which enables the use of many existing computational methods for simulation and data integration. To demonstrate, we apply our model integration approach to techno-social models of insurgent-led attacks and security force counter-measures centered on improvised explosive devices.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

Optimal experimental design: Formulations and computations

Questions of ‘how best to acquire data’ are essential to modelling and prediction in the natural and social sciences, engineering applications, and beyond. Optimal experimental design (OED) formalizes these questions and creates computational methods to answer them. This article presents a systematic survey of modern OED, from its foundations in classical design theory to current research involving OED for complex models. We begin by reviewing criteria used to formulate an OED problem and thus to encode the goal of performing an experiment. We emphasize the flexibility of the Bayesian and decision-theoretic approach, which encompasses information-based criteria that are well-suited to nonlinear and non-Gaussian statistical models. We then discuss methods for estimating or bounding the values of these design criteria; this endeavour can be quite challenging due to strong nonlinearities, high parameter dimension, large per-sample costs, or settings where the model is implicit. A complementary set of computational issues involves optimization methods used to find a design; we discuss such methods in the discrete (combinatorial) setting of observation selection and in settings where an exact design can be continuously parametrized. Finally we present emerging methods for sequential OED that build non-myopic design policies, rather than explicit designs; these methods naturally adapt to the outcomes of past experiments in proposing new experiments, while seeking coordination among all experiments to be performed. Throughout, we highlight important open questions and challenges.

97 MATHEMATICS AND COMPUTING↗

A Bayesian approach to time-domain photonic Doppler velocimetry analysis

Photonic Doppler velocimetry (PDV) is an established technique for measuring the velocities of fast-moving surfaces in high-energy-density experiments. In the standard approach to PDV analysis, the short-time Fourier transform (STFT) is used to generate a spectrogram from which the velocity history of the target is inferred. The user chooses the form, duration, and separation of the window function. Here, in this study, we present a Bayesian approach to infer the velocity directly from the PDV oscilloscope trace, without using the spectrogram for analysis. This is clearly a difficult inference problem due to the highly periodic nature of the data, but we find that with carefully chosen prior distributions for the model parameters, we can accurately recover the injected velocity from synthetic data. We validate this method using PDV data collected at the STAR two-stage light gas gun at Sandia National Laboratories, recovering shock-front velocities in quartz that are consistent with those inferred using the STFT-based approach and are interpolated across regions of low signal-to-noise data. Although this method does not rely on the same user choices as the STFT, we caution that it can be prone to misspecification if the chosen model is not sufficient to capture the velocity behavior. Analysis using posterior predictive checks can be used to establish whether a better model is required, although more complex models come with additional computational cost, often taking more than several hours to converge when sampling the Bayesian posterior. We, therefore, recommend it be viewed as a complementary method to that of the STFT-based approach.

Allison, James R. [First Light Fusion Ltd., Yarnto↗

Forensic Signature Detection of Yersinia Pestis Culturing Practices Across Institutions Using a Bayesian Network

The field of bioforensics is focused on the analysis of evidence from a biocrime. Existing laboratory analyses can identify the specific strain of an organism in the evidence, as well signatures of the specific culture batch of organisms, such as low-frequency contaminants or indicators of growth and processing methods. To link these disparate types of physical data to potential suspects, investigators may need to identify institutions or individuals whose access to strains and culturing practices match those identified from the evidence. In this work we present a Bayesian statistical network to fuse different types of analytical measurements that predict the production environment of a Yersinia pestis sample under investigation with automated test processing of scientific publications to identify institutions with a history of growing Y. pestis under similar conditions. Furthermore, the textual and experimental signatures were evaluated recursively to determine the overall sensitivity of the network across all levels of false positives. We illustrate that institutions associated with several specific culturing practices can be accurately selected based on the experimental signature from only a few analytical measurements. These findings demonstrate that similar Bayesian networks can be generated generically for many organisms of interest and their deployment is not prohibitive due to either computational or experimental factors.

Bayesian networks↗

Analytical methods for superresolution dislocation identification in dark-field X-ray microscopy

In this work, we develop several inference methods to estimate the position of dislocations from images generated using dark-field X-ray microscopy (DFXM)—achieving superresolution accuracy and principled uncertainty quantification. Using the framework of Bayesian inference, we incorporate models of the DFXM contrast mechanism and detector measurement noise, along with initial position estimates, into a statistical model coupling DFXM images with the dislocation position of interest. We motivate several position estimation and uncertainty quantification algorithms based on this model. We then demonstrate the accuracy of our primary estimation algorithm on synthetic realistic DFXM images of edge dislocations in single-crystal aluminum. We conclude with a discussion of our methods’ impact on future dislocation studies and possible future research avenues.

36 MATERIALS SCIENCE↗

Hierarchical Inference of the Lensing Convergence from Photometric Catalogs with Bayesian Graph Neural Networks

Abstract We present a Bayesian graph neural network (BGNN) that can estimate the weak lensing convergence ( κ ) from photometric measurements of galaxies along a given line of sight (LOS). The method is of particular interest in strong gravitational time-delay cosmography (TDC), where characterizing the “external convergence” ( κ ext ) from the lens environment and LOS is necessary for precise Hubble constant ( H 0 ) inference. Starting from a large-scale simulation with a κ resolution of ∼1′, we introduce fluctuations on galaxy–galaxy lensing scales of ∼1″ and extract random sight lines to train our BGNN. We then evaluate the model on test sets with varying degrees of overlap with the training distribution. For each test set of 1000 sight lines, the BGNN infers the individual κ posteriors, which we combine in a hierarchical Bayesian model to yield constraints on the hyperparameters governing the population. For a test field well sampled by the training set, the BGNN recovers the population mean of κ precisely and without bias (within the 2 σ credible interval), resulting in a contribution to the H 0 error budget well under 1%. In the tails of the training set with sparse samples, the BGNN, which can ingest all available information about each sight line, extracts a stronger κ signal compared to a simplified version of the traditional method based on matching galaxy number counts, which is limited by sample variance. Our hierarchical inference pipeline using BGNNs promises to improve the κ ext characterization for precision TDC. The code is available as a public Python package, Node to Joy ⏬ .

79 ASTRONOMY AND ASTROPHYSICS↗

Statistical tools for a better optical model

Background: Modern statistical tools provide the ability to compare the information content of observables and provide a path to explore which experiments would be most useful to give insight into and constrain theoretical models. Purpose: Here we study three such tools in the context of nuclear reactions with the goal of constraining the optical potential. Method: The three statistical tools examined are (i) the principal component analysis, (ii) the sensitivity analysis based on derivatives, and (iii) the Bayesian evidence. We first apply these tools to a toy-model case, comparing the form of the imaginary part of the optical potential. Then we consider two different reaction observables, elastic angular distributions and polarization data for reactions on 48 Ca and 208 Pb at two different beam energies. Results: For the toy-model case, we find significant discrimination power in the sensitivities and the Bayesian evidence, showing clearly that the volume imaginary term is more useful to describe scattering at higher energies. When comparing between elastic cross sections and polarization data using realistic optical models, sensitivity studies indicate that both observables are roughly equally sensitive but the variability of the optical model parameters is strongly angle dependent. The Bayesian evidence shows some variability between the two observables, but the Bayes factor obtained is not sufficient to discriminate between angular distributions and polarization. Conclusions: From the cases considered, we conclude that, in general, elastic scattering angular distributions have similar impact in constraining the optical potential parameters compared with the polarization data. The angular ranges for the optimum experimental constraints can vary significantly with the observable considered.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Towards Improving luminosity using optics tuning and data-driven methods

The results of Run 24 experiments at Relativistic Heavy Ion Collider (RHIC) for improving luminosity using optics tuning are presented in this study. In the first experiment, MADx matching was used to output magnet strengths corresponding to specific s star movements around Interaction Region 8 (IR8). The corresponding Zero Degree Calorimeter (ZDC) signal was measured in place of luminosity, and Bayesian Optimization aids search of optimal movements. It was found that values retrieved from matching were inaccurate, resulting in negative feedback loops. The second experiment focused on calculating accurate s star movements. The matching method was replaced with a linear sensitivity matrix, directly relating optics to power supply, and its null space was used to fit constraints such as hysteresis effects. At the experiment, beam losses were observed at collimators around boundary of IR8, which were fixed for the third experiment. Dynamic mode decomposition was also introduced to improve quality of turn-by-turn (TBT) data as well as accuracy and consistency of optics measurements at IR8. These improvements will be tested in the experiment of next RHIC run for luminosity optimization.

Accelerator Physics↗

Distinguishing cause from correlation in tokamak experiments to trigger edge-localised plasma instabilities

The generic question is considered: How can we determine the probability of an otherwise quasi-random event, having been triggered by an external influence? A specific problem is the quantification of the success of techniques to trigger, and hence control, edge-localised plasma instabilities (ELMs) in magnetically confined fusion (MCF) experiments. The development of such techniques is essential to ensure tolerable heat loads on components in large MCF fusion devices, and is necessary for their development into economically successful power plants. Bayesian probability theory is used to rigorously formulate the problem and to provide a formal solution. Accurate but pragmatic methods are developed to estimate triggering probabilities, and are illustrated with experimental data. These allow results from experiments to be quantitatively assessed, and rigorously quantified conclusions to be formed. Example applications include assessing whether triggering of ELMs is a statistical or deterministic process, and the establishment of thresholds to ensure that ELMs are reliably triggered.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DISSECTING THE GRAVITATIONAL LENS B1608+656. II. PRECISION MEASUREMENTS OF THE HUBBLE CONSTANT, SPATIAL CURVATURE, AND THE DARK ENERGY EQUATION OF STATE

Strong gravitational lens systems with measured time delays between the multiple images provide a method for measuring the 'time-delay distance' to the lens, and thus the Hubble constant. We present a Bayesian analysis of the strong gravitational lens system B1608+656, incorporating (1) new, deep Hubble Space Telescope (HST) observations, (2) a new velocity-dispersion measurement of 260 +- 15 km s{sup -1} for the primary lens galaxy, and (3) an updated study of the lens' environment. Our analysis of the HST images takes into account the extended source surface brightness, and the dust extinction and optical emission by the interacting lens galaxies. When modeling the stellar dynamics of the primary lens galaxy, the lensing effect, and the environment of the lens, we explicitly include the total mass distribution profile logarithmic slope gamma' and the external convergence kappa{sub ext}; we marginalize over these parameters, assigning well-motivated priors for them, and so turn the major systematic errors into statistical ones. The HST images provide one such prior, constraining the lens mass density profile logarithmic slope to be gamma' = 2.08 +- 0.03; a combination of numerical simulations and photometric observations of the B1608+656 field provides an estimate of the prior for kappa{sub ext}: 0.10{sup +0.08}{sub -0.05}. This latter distribution dominates the final uncertainty on H{sub 0}. Fixing the cosmological parameters at OMEGA{sub m} = 0.3, OMEGA{sub L}AMBDA = 0.7, and w = -1 in order to compare with previous work on this system, we find H{sub 0} = 70.6{sup +3.1}{sub -3.1} km s{sup -1} Mpc{sup -1}. The new data provide an increase in precision of more than a factor of 2, even including the marginalization over kappa{sub ext}. Relaxing the prior probability density function for the cosmological parameters to that derived from the Wilkinson Microwave Anisotropy Probe (WMAP) five-year data set, we find that the B1608+656 data set breaks the degeneracy between OMEGA{sub m} and OMEGA{sub L}AMBDA at w = -1 and constrains the curvature parameter to be -0.031 < OMEGA{sub k} < 0.009 (95% CL), a level of precision comparable to that afforded by the current Type Ia SNe sample. Asserting a flat spatial geometry, we find that, in combination with WMAP, H{sub 0} = 69.7{sup +4.9}{sub -5.0} km s{sup -1} Mpc{sup -1} and w = -0.94{sup +0.17}{sub -0.19} (68% CL), suggesting that the observations of B1608+656 constrain w as tightly as the current Baryon Acoustic Oscillation data do.

79 ASTRONOMY AND ASTROPHYSICS↗

Development of Bayesian analysis program for extraction of polarisation observables at CLAS

At the mass scale of a proton, the strong force is not well understood. Various quark models exist, but it is important to determine which quark model(s) are most accurate. Experimentally, finding resonances predicted by some models and not others would give valuable insight into this fundamental interaction. Several labs around the world use photoproduction experiments to find these missing resonances. The aim of this work is to develop a robust Bayesian data analysis program for extracting polarisation observables from pseudoscalar meson photoproduction experiments using CLAS at Jefferson Lab. This method, known as nested sampling, has been compared to traditional methods and has incorporated data parallelisation and GPU programming. It involves an event-by-event likelihood function, which has no associated loss of information from histogram binning, and results can be easily constrained to the physical region. One of the most important advantages of the nested sampling approach is that data from different experiments can be combined and analysed simultaneously. Results on both simulated and previously analysed experimental data for the K{sup +}Λ channel will be discussed.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

A Copula-Based Conditional Probabilistic Forecast Model for Wind Power Ramps

Efficient management of wind ramping characteristics can significantly reduce wind integration costs for balancing authorities. By considering the stochastic dependence of wind power ramp (WPR) features, this paper develops a conditional probabilistic wind power ramp forecast (cp-WPRF) model based on Copula theory. The WPRs dataset is constructed by extracting ramps from a large dataset of historical wind power. Each WPR feature (e.g., rate, magnitude, duration, and start-time) is separately forecasted by considering the coupling effects among different ramp features. To accurately model the marginal distributions with a copula, a Gaussian mixture model (GMM) is adopted to characterize the WPR uncertainty and features. The Canonical Maximum Likelihood (CML) method is used to estimate parameters of the multivariable copula. The optimal copula model is chosen based on the Bayesian information criterion (BIC) from each copula family. Finally, the best conditions based cp-WPRF model is determined by predictive interval (PI) based evaluation metrics. Numerical simulations on publicly available wind power data show that the developed copula-based cp-WPRF model can predict WPRs with a high level of reliability and sharpness.

17 WIND ENERGY↗

Efficient compressive and Bayesian characterization of biphoton frequency spectra

Frequency-bin qudits constitute a promising tool for quantum information processing, but their high dimensionality can make for tedious characterization measurements. Here we introduce and compare compressive sensing and Bayesian mean estimation for recovering the spectral correlations of entangled photon pairs. Using a conventional compressive sensing algorithm, we reconstruct joint spectra with up to a 26-fold reduction in measurement time compared to the equivalent raster scan. Applying a custom Bayesian model to the same data, we then additionally realize reliable and consistent quantification of uncertainty. Finally, these efficient methods of biphoton characterization should advance our ability to use the high degree of parallelism and complexity afforded by frequency-bin encoding.

42 ENGINEERING↗

Large-Scale Optimization for Bayesian Inference in Complex Systems

The SAGUARO (Scalable Algorithms for Groundwater Uncertainty Analysis and Robust Optimization) Project focused on the development of scalable numerical algorithms for large-scale Bayesian inversion in complex systems that capitalize on advances in large-scale simulation-based optimization and inversion methods. The project was a collaborative effort among MIT, the University of Texas at Austin, Georgia Institute of Technology, and Sandia National Laboratories. The research was directed in three complementary areas: efficient approximations of the Hessian operator, reductions in complexity of forward simulations via stochastic spectral approximations and model reduction, and employing large-scale optimization concepts to accelerate sampling. The MIT₋Sandia component of the SAGUARO Project addressed the intractability of conventional sampling methods for large-scale statistical inverse problems by devising reduced-order models that are faithful to the full-order model over a wide range of parameter values; sampling then employs the reduced model rather than the full model, resulting in very large computational savings. Results indicate little effect on the computed posterior distribution. On the other hand, in the Texas₋Georgia Tech component of the project, we retain the full-order model, but exploit inverse problem structure (adjoint-based gradients and partial Hessian information of the parameter-to-observation map) to implicitly extract lower dimensional information on the posterior distribution; this greatly speeds up sampling methods, so that fewer sampling points are needed. We can think of these two approaches as "reduce then sample" and "sample then reduce." In fact, these two approaches are complementary, and can be used in conjunction with each other. Moreover, they both exploit deterministic inverse problem structure, in the form of adjoint-based gradient and Hessian information of the underlying parameter-to-observation map, to achieve their speedups.

97 MATHEMATICS AND COMPUTING↗

Improving and Assessing the Quality of Uncertainty Quantification in Deep Learning

Deep learning (DL) models have enjoyed increased attention in recent years because of their powerful predictive capabilities. While many successes have been achieved, standard deep learning methods suffer from a lack of uncertainty quantification (UQ). While the development of methods for producing UQ from DL models is an active area of current research, little attention has been given to the quality of the UQ produced by such methods. In order to deploy DL models to high-consequence applications, high-quality UQ is necessary. This report details the research and development conducted as part of a Laboratory Directed Research and Development (LDRD) project at Sandia National Laboratories. The focus of this project is to develop a framework of methods and metrics for the principled assessment of UQ quality in DL models. This report presents an overview of UQ quality assessment in traditional statistical modeling and describes why this approach is difficult to apply in DL contexts. An assessment on relatively simple simulated data is presented to demonstrate that UQ quality can differ greatly between DL models trained on the same data. A method for simulating image data that can then be used for UQ quality assessment is described. A general method for simulating realistic data for the purpose of assessing a model’s UQ quality is also presented. A Bayesian uncertainty framework for understanding uncertainty and existing metrics is described. Research that came out of collaborations with two university partners are discussed along with a software toolkit that is currently being developed to implement the UQ quality assessment framework as well as serve as a general guide to incorporating UQ into DL applications.

97 MATHEMATICS AND COMPUTING↗