Search NASA⌕ Search

SEARCH · Search NASA

Results for “Model selection”

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 487 records · Page 27

An Optimal Parameterization Framework for Infrasonic Tomography of the Stratospheric Winds Using Non-Local Sources

A method is developed to apply acoustic tomography methods to a localized network of infrasound arrays with intention of monitoring the atmosphere state in the region around the network using non-local sources without requiring knowledge of the precise source location or non-local atmosphere state. Closely spaced arrays provide a means to estimate phase velocities of signals that can provide limiting bounds on certain characteristics of the atmosphere. Larger spacing between such clusters provide a means to estimate celerity from propagation times along multiple unique stratospherically or thermospherically ducted propagation paths and compute more precise estimates of the atmosphere state. In order to avoid the commonly encountered complex, multimodal distributions for parametric atmosphere descriptions and to maximize the computational efficiency of the method, an optimal parametrization framework is constructed. This framework identifies the ideal combination of parameters for tomography studies in specific regions of the atmosphere and statistical model selection analysis shows that high quality corrections to the middle atmosphere winds can be obtained using as few as three parameters. Lastly, comparison of the resulting estimates for synthetic data sets shows qualitative agreement between the middle atmosphere winds and those estimated from infrasonic traveltime observations.

58 GEOSCIENCES↗

A deep, multi-epoch Chandra HETG study of the ionized outflow from NGC 4051

Actively accreting supermassive black holes significantly impact the evolution of their host galaxies, truncating further star formation by expelling large fractions of gas with wide-angle outflows. The X-ray band is key to understanding how these black hole winds affect their environment, as the outflows have high temperatures ~10 5–8 K). We have developed a Bayesian framework for characterizing active galactic nucleus outflows with an improved ability to explore parameter space and perform robust model selection. We applied this framework to a new 700 ks and an archival 315 ks Chandra High Energy Transmission Gratings observation of the Seyfert galaxy NGC 4051. We have detected six absorbers intrinsic to NGC 4051. These wind components span velocities from 400 to 30000 km s -1 . We have determined that the most statistically significant wind component is purely collisionally ionized, which is the first detection of such an absorber. This wind has T ≈ 10 7 K and v ≈ 880 km s -1 and remains remarkably stable between the two epochs. Other slow components also remain stable across time. Fast outflow components change their properties between 2008 and 2016, suggesting either physical changes or clouds moving in and out of the line of sight. For one of the fast components, we obtain one of the tightest wind density measurements to date, log n/(cm -3 ) = 13.0$^{+0.01}_{-0.02}$, and determine that it is located at ~240 gravitational radii. The estimated total outflow power surpasses 5 percent of the bolometric luminosity (albeit with large uncertainties) making it important in the context of galaxy–black hole interactions.

79 ASTRONOMY AND ASTROPHYSICS↗

Photometric redshift uncertainties in weak gravitational lensing shear analysis: models and marginalization

ABSTRACT Recovering credible cosmological parameter constraints in a weak lensing shear analysis requires an accurate model that can be used to marginalize over nuisance parameters describing potential sources of systematic uncertainty, such as the uncertainties on the sample redshift distribution n(z). Due to the challenge of running Markov chain Monte Carlo (MCMC) in the high-dimensional parameter spaces in which the n(z) uncertainties may be parametrized, it is common practice to simplify the n(z) parametrization or combine MCMC chains that each have a fixed n(z) resampled from the n(z) uncertainties. In this work, we propose a statistically principled Bayesian resampling approach for marginalizing over the n(z) uncertainty using multiple MCMC chains. We self-consistently compare the new method to existing ones from the literature in the context of a forecasted cosmic shear analysis for the HSC three-year shape catalogue, and find that these methods recover statistically consistent error bars for the cosmological parameter constraints for predicted HSC three-year analysis, implying that using the most computationally efficient of the approaches is appropriate. However, we find that for data sets with the constraining power of the full HSC survey data set (and, by implication, those upcoming surveys with even tighter constraints), the choice of method for marginalizing over n(z) uncertainty among the several methods from the literature may modify the 1σ uncertainties on Ωm–S8 constraints by ∼4 per cent, and a careful model selection is needed to ensure credible parameter intervals.

Zhang, Tianqing (ORCID:000000025596198X)↗

Application of deep learning methods for beam size control during user operation at the Advanced Light Source

Past research at the Advanced Light Source (ALS) provided a proof-of-principle demonstration that deep learning methods could be effectively employed to compensate for the significant perturbations to the transverse electron beam size induced by user-controlled adjustments of the insertion devices. However, incorporating these methods into the ALS’ daily operations has faced notable challenges. The complexity of the system’s operational requirements and the significant upkeep demands has restricted their sustained application during user operation. Here, we introduce the development of a more robust neural network (NN)-based algorithm that utilizes a novel online fine-tuning approach and its systematic integration into the day-to-day machine operations. Our analysis emphasizes the process of NN model selection, demonstrates the superior performance of the NN-based method over traditional feedback methods, and examines the effectiveness and resilience of the new algorithm during user-operation scenarios. Published by the American Physical Society 2024

43 PARTICLE ACCELERATORS↗

Fast and accurate simulations of calorimeter showers with normalizing flows

In this study, we introduce caloflow, a fast detector simulation framework based on normalizing flows. For the first time, we demonstrate that normalizing flows can reproduce many-channel calorimeter showers with extremely high fidelity, providing a fresh alternative to computationally expensive geant4 simulations, as well as other state-of-the-art fast simulation frameworks based on generative adversarial networks (GANs) or variational autoencoders (VAEs). In addition to the usual histograms of physical features and images of calorimeter showers, we introduce a new metric for judging the quality of generative modeling: the performance of a classifier trained to differentiate real from generated images. We show that GAN-generated images can be identified by the classifier with nearly 100% accuracy, while images generated from caloflow are better able to fool the classifier. More broadly, normalizing flows offer several advantages compared to other state-of-the-art approaches (GANs and VAEs), including tractable likelihoods, stable and convergent training, and principled model selection. Normalizing flows also provide a bijective mapping between data and the latent space, which could have other applications beyond simulation, for example, to detector unfolding.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Towards the minimal spectrum of excited baryons

In the light baryon sector the resonances can be broad and overlapping and are in most cases not directly visible in the cross section data. Automatized model selection techniques that introduce penalties for choosing particular resonance sets can be used to determine the minimally needed set of resonances to describe the data. Several possible penalization schemes for the occurrence of resonances are compared. As an application we perform a blindfold identification of hyperon resonances in the $\bar{K}$ N → K $\Xi$ reaction based on the Least Absolute Shrinkage and Selection Operator (LASSO) in combination with the Bayesian Information Criterion (BIC). Here, we find ten resonances - out of the 21 above-threshold hyperon resonances with spin J ≤ 7/2 listed by the Particle Data Group. In traditional analyses, it is practically impossible to test the relevance of all resonances and their combinations that may potentially contribute to the reaction. By contrast, the present method proves capable of determining the relevant resonances among a relatively large pool of candidates.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Cosmic distance inference from purely geometric BAO methods: Linear point standard ruler and correlation function model fitting

Leveraging the baryon acoustic oscillations (BAO) feature present in clustering 2-point statistics, we aim to measure cosmological distances independently of the underlying background cosmological model. However this inference is complicated by late-time nonlinearities that introduce model and tracer dependencies in the clustering correlation function and power spectrum, which must be properly accounted for. With this in mind, we introduce the “purely geometric-BAO,” which provides a rigorous tool to measure cosmological distances without assuming a specific background cosmology. We focus on the 2-point clustering correlation function monopole, and show how to implement such an inference scheme employing two different methodologies: the linear point standard ruler (LP) and correlation-function model-fitting (CF-MF). For the first time we demonstrate how, by means of the CF-MF, we can measure very precisely the sound-horizon/isotropic-volume-distance ratio, r d /DV(z¯), while correctly propagating all the uncertainties. Using synthetic data, we compare the outcomes of the two methodologies, and find that the LP provides up to 50% more precise measurements than the CF-MF. Finally, we test a procedure widely employed in BAO analyses: fitting the 2-point function while fixing the cosmological and the non-linear-damping parameters at fiducial values. We find that this underestimates the distance errors by nearly a factor of 2. Furthermore, we recommend that this practice be reconsidered, whether for parameter determination or model selection.

79 ASTRONOMY AND ASTROPHYSICS↗

Footprints of the QCD Crossover on Cosmological Gravitational Waves at Pulsar Timing Arrays

Pulsar timing arrays (PTAs) have reported evidence for a stochastic gravitational wave (GW) background at nanohertz frequencies, possibly originating in the early Universe. We show that the spectral shape of the low-frequency (causality) tail of GW signals sourced at temperatures around T ≳ 1 GeV is distinctively affected by confinement of strong interactions (QCD), due to the corresponding sharp decrease in the number of relativistic species, and significantly deviates from ∼ f 3 commonly adopted in the literature. Bayesian analyses in the NANOGrav 15 years and the previous international PTA datasets reveal a significant improvement in the fit with respect to cubic power-law spectra, previously employed for the causality tail. While no conclusion on the nature of the signal can be drawn at the moment, our results show that the inclusion of standard model effects on cosmological GWs can have a decisive impact on model selection. Published by the American Physical Society 2024

Franciolini, Gabriele (ORCID:0000000268929145)↗

A Physics-Aligned Multi-Domain Machine Learning Framework for Time-Localised Diagnosis of Power Electronics Faults

This paper presents a physics-aligned framework for fault diagnosis in multi-phase power-electronic systems using cycle-synchronous windowing and multi-domain features derived from Fourier, wavelet, and Hilbert–Huang representations. While both logistic regression and multilayer perceptron (MLP) models achieve perfect performance under standard evaluation, blind unseen testing reveals a critical failure in a baseline MLP. This is shown to arise from model selection based on validation accuracy. Using validation-loss-based selection restores correct unseen performance and improves confidence. Feature ablation shows that Fourier and wavelet features dominate, while computational analysis indicates that feature extraction, particularly HHT, governs runtime.

Kumar, Praveen [ORNL] (ORCID:0000000291877857)↗

A Performance Model for Allocating the Parallelism in a Multigrid-in-Time Solver

The traditional way to numerically solve time-dependent problems is to sequentially march through time, solving for one time step and then the next. The parallelism in this approach is limited to the spatial dimension, which is quickly exhausted, causing the gain from using more processors to solve a problem to diminish. One approach to overcome this barrier is to use methods that are parallel in time. These methods have the potential to achieve dramatically better performance compared to time-stepping approaches, but achieving this performance requires carefully choosing the amount of parallelism devoted to space versus the amount devoted to time. Here, we present a performance model that, for a multigrid-in-time solver, makes the decision on when to switch to parallel-in-time and on how much parallelism to devote to space vs. time. In our experiments, the model selects the best parallel configuration in most of our test cases and a configuration close to the best one in all other cases.

97 MATHEMATICS AND COMPUTING↗

Event Detection and Classification Using Machine Learning Applied to PMU Data for the Western US Power System

Smart grid technology enhances our comprehension and reliability of the power grid, leveraging Phasor Measurement Unit (PMU) data—time-synchronized, high-frequency measurements gathered across the US power grid. This paper employs machine learning techniques to effectively analyze the vast PMU data in Wide Area Monitoring Systems (WAMS) for power grid event detection and classification. Analyzing several months of real-world PMU data, the paper focuses on machine learning for fast, precise event detection and classification, corroborated by utility event logs. Practical challenges like feature extraction, dimensionality reduction, and model selection are addressed. A novel feature yielding improved results is discovered, and a supplementary algorithm for detecting small power grid faults is developed. The final algorithm is validated using a month-long real PMU data set, demonstrating its capability in accurately identifying power grid events in near real-time.

machine learning, event detection, PMU↗

Spread of an introduced parasite across the Hawaiian archipelago independent of its introduced host

1. Co-introductions of non-native parasites with non-native hosts can be a major driver of disease emergence in native species, but the conditions that promote the establishment and spread of nonnative parasites remain poorly understood. Here, we characterise the infection of a native host species by a non-native parasite relative to the distribution and density of the original non-native host species and a suite of organismal and environmental factors that have been associated with parasitism, but not commonly considered within a single system. 2. We examined the native Hawaiian goby Awaous stamineus across 23 catchments on five islands for infection by the non-native nematode parasite Camallanus cotti . We used model selection to test whether parasite infection was associated with the genetic diversity, size and population density of native hosts, the distribution and density of non-native hosts, land use and water quality. 3. We found that the distribution of non-native C. cotti parasites has become decoupled from the non-native hosts that were primary vectors of introduction to the Hawaiian Islands. Although no single intrinsic or extrinsic factor was identified that best explains parasitism of A. stamineus by C. cotti , native host size, population density and water quality were consistently identified as influencing parasite intensity and prevalence. 4. The introduction of non-native species can indirectly influence native species through infection of co-introduced parasites. Here, we show that the effects of enemy addition can extend beyond the range of non-native hosts through the independent spread of non-native parasites. This suggests that control of non-native hosts is not sufficient to halt the spread of introduced parasites. Furthermore, designing importation regulations to prevent host parasite co-introductions can promote native species conservation, even in remote areas that may not seem susceptible to human influence.

59 BASIC BIOLOGICAL SCIENCES↗

Rapid Reflective Facet Characterization Using Fringe Reflection Techniques

Reflective mirror facets for concentrating solar power (CSP) systems have stringent requirements on the surface slope accuracy in order to provide adequate system performance. This paper presents a tool that can fully characterize facets quickly enough for 100% inspection on a production line. A facet for a CSP system, specifically a dish concentrator has a parabolic design shape. This shape will concentrate near-parallel rays from the sun to a point (or a line for trough systems). Deviations of surface slope from the design shape impact the performance of the system, either losing power that misses the target or increasing peak fluxes to undesirable levels. During development or production, accurate knowledge of facet defects can lead to improvements to lower cost or improve performance. The reported characterization system, SOFAST (Sandia Optical Fringe Analysis Slope Tool), has a computer-connected camera that images the reflective surface, which is positioned so that it reflects an active target, such as an LCD screen, to the camera. A series of fringe patterns are displayed on the screen while images are captured. Using the captured information, the reflected target location of each pixel of mirror viewed can be determined, and thus through a mathematical transformation, a surface normal map can be developed. This is then fitted to the selected model equation, and the errors from design are characterized. While similar approaches have been explored, several key developments are presented here. The combination of the display, capture, and data reduction in one system allows rapid characterization. An “electronic boresight” approach is utilized to accommodate physical equipment positioning deviations, making the system insensitive to setup errors. Up to 1.5×10 6 points are characterized on each facet. Finally, while prior automotive industry commercial systems resolve the data to shape determination, SOFAST concentrates on slope characterization and reporting, which is tailored to solar applications. SOFAST can be used for facet analysis during development. However, the real payoff is in production, where complete analysis is performed in about 10s. With optimized coding, this could be further reduced.

14 SOLAR ENERGY↗

Comparison of Radiation Models for a Turbulent Piloted Methane/Air Jet Flame: A Frozen-Field Study

Numerical modeling of radiative transfer in nongray reacting media is a challenging problem in computational science and engineering. The choice of radiation models is important for accurate and efficient high-fidelity combustion simulations. Different applications usually involve different degrees of complexity, so there is yet no consensus in the community. In this paper, the performance of different radiative transfer equation (RTE) solvers and spectral models for a turbulent piloted methane/air jet flame are studied. The flame is scaled from the Sandia Flame D with a Reynolds number of 22,400. Three classes of RTE solvers, namely the discrete ordinates method, spherical harmonics method, and Monte Carlo method, are examined. The spectral models include the Planck-mean model, the full-spectrum k-distribution (FSK) method, and the line-by-line (LBL) calculation. The performances of different radiation models in terms of accuracy and computational cost are benchmarked. The results have shown that both RTE solvers and spectral models are critical in the prediction of radiative heat source terms for this jet flame. The trade-offs between the accuracy, the computational cost, and the implementation difficulty are discussed in detail. The results can be used as a reference for radiation model selection in combustor simulations.

David, Chloe↗

Growth of FePt encapsulated carbon nanotubes by thermal chemical vapor deposition

FePt encapsulated carbon nanotubes (CNTs) were grown by thermal chemical vapor deposition using an Fe/Pt bilayer catalyst. The CNTs were grown according to the base growth model. Selected area electron diffraction results revealed that the encapsulated particles were A1-FePt, L1{sub 0}-FePt, and Fe{sub 3}PtC. The crystal structures of particles found at the root parts of CNTs were not able to be identified, however. The layered structure of catalytic films seemed to be responsible for the difference in Pt content between particles found at tip and root parts of CNTs. Approximately 60% of CNTs grown at 800 °C had particles at their tip parts, compared to only 30% when the growth temperature was 700 °C, indicating that higher process temperatures promote particle encapsulation in CNTs.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Properties of Optical Ceramics CO1 and CO2 upon Modification of Their Surface by Carbon Nanotubes

The results of a study of the modification of the properties of optical inorganic materials (optical ceramics CO1 (MgF{sub 2}) and CO2 (ZnS)) are presented. It is found that the main characteristics of the studied materials with a surface modified by carbon nanotubes significantly changes when the laser-oriented deposition method is applied. Main features of the carbon nanotubes caused by its developed surface, the strength of C–C bonds, and low value of refraction index are taken into consideration. Analytical, quantum-chemical and experimental results of studies of changes in the spectral and mechanical parameters, as well as the wetting angle of the selected model inorganic matrices, are given.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

A Variational Inference Approach to Inverse Problems with Gamma Hyperpriors

Hierarchical models with gamma hyperpriors provide a flexible, sparse-promoting framework to bridge L 1 and L 2 regularizations in Bayesian formulations to inverse problems. Despite the Bayesian motivation for these models, existing methodologies are limited to maximum a posteriori estimation. The potential to perform uncertainty quantification has not yet been realized. This paper introduces a variational iterative alternating scheme for hierarchical inverse problems with gamma hyperpriors. The proposed variational inference approach yields accurate reconstruction, provides meaningful uncertainty quantification, and is easy to implement. In addition, it lends itself naturally to conduct model selection for the choice of hyperparameters. Here, we illustrate the performance of our methodology in several computed examples, including a deconvolution problem and sparse identification of dynamical systems from time series data.

Bayesian shrinkage↗

Optimizing deep learning hyper-parameters through an evolutionary algorithm. In: MLHPC '15 Proceedings of the Workshop on Machine Learning in High-Performance Computing Environments, Article No. 4

There has been a recent surge of success in utilizing Deep Learning (DL) in imaging and speech applications for its relatively automatic feature generation and, in particular for convolutional neural networks (CNNs), high accuracy classification abilities. While these models learn their parameters through data-driven methods, model selection (as architecture construction) through hyper-parameter choices remains a tedious and highly intuition driven task. To address this, Multi-node Evolutionary Neural Networks for Deep Learning (MENNDL) is proposed as a method for automating network selection on computational clusters through hyper-parameter optimization performed via genetic algorithms.

Young, Steven R.↗