Search NASA⌕ Search

SEARCH · Search NASA

Results for “Bayesian method”

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 1,171 records · Page 65

Spline Classification Methods

The use of spline functions in the development of classification algorithms is discussed. A method is formulated for producing spline approximations to univariate density functions when each density function is described by a histogram of measurements. The resulting approximations are then incorporated into a Bayesian classification procedure for which the probability of misclassification can be readily computed. Some preliminary numerical results are presented to illustrate the method.

Guseman, L. F., Jr.↗

Adaptive hyperparameter updating for training restricted Boltzmann machines on quantum annealers

Restricted Boltzmann Machines (RBMs) have been proposed for developing neural networks for a variety of unsupervised machine learning applications such as image recognition, drug discovery, and materials design. The Boltzmann probability distribution is used as a model to identify network parameters by optimizing the likelihood of predicting an output given hidden states trained on available data. Training such networks often requires sampling over a large probability space that must be approximated during gradient based optimization. Quantum annealing has been proposed as a means to search this space more efficiently which has been experimentally investigated on D-Wave hardware. D-Wave implementation requires selection of an effective inverse temperature or hyperparameter (β) within the Boltzmann distribution which can strongly influence optimization. Here, we show how this parameter can be estimated as a hyperparameter applied to D-Wave hardware during neural network training by maximizing the likelihood or minimizing the Shannon entropy. We find both methods improve training RBMs based upon D-Wave hardware experimental validation on an image recognition problem. Neural network image reconstruction errors are evaluated using Bayesian uncertainty analysis which illustrate more than an order magnitude lower image reconstruction error using the maximum likelihood over manually optimizing the hyperparameter. The maximum likelihood method is also shown to out-perform minimizing the Shannon entropy for image reconstruction.

97 MATHEMATICS AND COMPUTING↗

Demonstration of an AI-driven workflow for dynamic x-ray spectroscopy

X-ray absorption near edge structure (XANES) spectroscopy is a powerful technique for characterizing the chemical state and symmetry of individual elements within materials, but requires collecting data at many energy points which can be time-consuming. While adaptive sampling methods exist for efficiently collecting spectroscopic data, they often lack domain-specific knowledge about the structure of XANES spectra. Here we demonstrate a knowledge-injected Bayesian optimization approach for adaptive XANES data collection that incorporates understanding of spectral features like absorption edges and pre-edge peaks. We show this method accurately reconstructs the absorption edge of XANES spectra using only 15–20% of the measurement points typically needed for conventional sampling, while maintaining the ability to determine the x-ray energy of the sharp peak after the absorption edge with errors less than 0.03 eV, the absorption edge with errors less than 0.1 eV; and overall root-mean-square errors less than 0.005 compared to traditionally sampled spectra. Our experiments on battery materials and catalysts demonstrate the method’s effectiveness for both static and dynamic XANES measurements, improving data collection efficiency and enabling better time resolution for tracking chemical changes. This approach advances the degree of automation in XANES experiments, reducing the common errors of under- or over-sampling points near the absorption edge and enabling dynamic experiments that require high temporal resolution or limited measurement time.

Bayesian optimization↗

Signal discovery, limits, and uncertainties with sparse on/off measurements: an objective bayesian analysis

For decades researchers have studied the On/Off counting problem where a measured rate consists of two parts. One part is due to a signal process and the other is due to a background process, the magnitudes for both of which are unknown. While most frequentist methods are adequate for large number counts, they cannot be applied to sparse data. Here, I want to present a new objective Bayesian solution that only depends on three parameters: the number of events in the signal region, the number of events in the background region, and the ratio of the exposure for both regions. First, the probability of the counts only being due to background is derived analytically. Second, the marginalized posterior for the signal parameter is also derived analytically. With this two-step approach it is easy to calculate the signal's significance, strength, uncertainty, or upper limit in a unified way. This approach is valid without restrictions for any number count, including zero, and may be widely applied in particle physics, cosmic-ray physics, and high-energy astrophysics. In order to demonstrate the performance of this approach, I apply the method to gamma-ray burst data.

79 ASTRONOMY AND ASTROPHYSICS↗

Bayesian prior construction for uncertainty quantification in first-principles statistical mechanics

First-principles statistical mechanics enables the prediction of thermodynamic and kinetic properties of materials, but is computationally expensive. Many approaches require surrogate models to calculate energies within Monte Carlo or molecular dynamics simulations. Inexpensive surrogates such as cluster expansions enable otherwise intractable calculations by interpolating data from higher accuracy methods, such as Density Functional Theory (DFT). Surrogate models introduce uncertainty into downstream calculations, in addition to any uncertainty inherent to DFT calculations. Bayesian frameworks address this by quantifying uncertainty and incorporating expert knowledge through priors. However, constructing effective priors remains challenging. This work introduces and describes practical strategies for building Bayesian cluster expansions, focusing on basis truncation, hyperparameter selection, and ground state replication. We analyze multiple basis truncation schemes, compare cross-validation to the evidence-approximation for hyperparameter optimization, and provide methods to find and enforce ground-state-preserving models through priors. Additionally, we compare the uncertainties between different approximations to DFT (LDA, PBE, SCAN) against the uncertainty introduced with the use of cluster expansion surrogate models. These approaches are demonstrated on the BCC Li x Mg 1-x and Li x Al 1-x alloys, which are both of interest for solid-state Li batteries. Our results provide guidelines for constructing and utilizing Bayesian cluster expansions, thereby improving the transparency of materials modeling. Furthermore, the approaches and insights developed in this work can be transferred to a wide range of cluster expansion surrogate models, including the atomic cluster expansion and related machine-learned interatomic potential architectures.

Alloy theory↗

Domain Knowledge Guided Bayesian Optimization For Autonomous Alignment Of Complex Scientific Instruments

Bayesian Optimization (BO) is a powerful tool for optimizing complex non-linear systems. However, its performance degrades in high-dimensional problems with tightly coupled parameters and highly asymmetric objective landscapes, where rewards are sparse. In such needle-in-a-haystack scenarios, even advanced methods like trust-region BO (TurBO) often lead to unsatisfactory results. We propose a domain knowledge guided Bayesian Optimization approach, which leverages physical insight to fundamentally simplify the search problem by transforming coordinates to decouple input features and align the active subspaces with the primary search axes. We demonstrate this approach's efficacy on a challenging 12-dimensional, 6-crystal Split-and-Delay optical system, where conventional approaches, including standard BO, TuRBO and multi-objective BO, consistently led to unsatisfactory results. When combined with an reverse annealing exploration strategy, this approach reliably converges to the global optimum. The coordinate transformation itself is the key to this success, significantly accelerating the search by aligning input co-ordinate axes with the problem's active subspaces. As increasingly complex scientific instruments, from large telescopes to new spectrometers at X-ray Free Electron Lasers are deployed, the demand for robust high-dimensional optimization grows. Our results demonstrate a generalizable paradigm: leveraging physical insight to transform high-dimensional, coupled optimization problems into simpler representations can enable rapid and robust automated tuning for consistent high performance while still retaining current optimization algorithms.

FOS: Computer and information sciences↗

RED CLUMP STARS IN THE SAGITTARIUS TIDAL STREAMS

We have probed a section (l {approx} 150, b {approx} -60) of the trailing tidal arm of the Sagittarius dwarf spheroidal galaxy by identifying a sample of Red Clump (RC) stream stars. RC stars are not generally found in the halo field, but are found in significant numbers in both the Sagittarius galaxy and its tidal streams, making them excellent probes of stream characteristics. Our target sample was selected using photometric data from the Sloan Digital Sky Survey, Data Release 6, which was constrained in color to match the Sagittarius RC stars. Spectroscopic observations of the target stars were conducted at Kitt Peak National Observatory using the WIYN telescope. The resulting spectroscopic sample is magnitude limited and contains both main-sequence disk stars and evolved RC stars. We have developed a method to systematically separate these two stellar classes using kinematic information and a Bayesian approach for surface gravity determination. The resulting RC sample allows us to determine an absolute stellar density of {rho} = 2.7 {+-} 0.5 RC stars kpc{sup -3} at this location in the stream. Future measurements of stellar densities for a variety of populations and at various locations along the streams will lead to a much improved understanding of the original nature of the Sagittarius galaxy and the physical processes controlling its disruption and subsequent stream generation.

79 ASTRONOMY AND ASTROPHYSICS↗

STELLAR POPULATIONS IN THE CENTRAL 0.5 pc OF THE GALAXY. I. A NEW METHOD FOR CONSTRUCTING LUMINOSITY FUNCTIONS AND SURFACE-DENSITY PROFILES

We present new high angular resolution near-infrared spectroscopic observations of the nuclear star cluster surrounding the Milky Way's central supermassive black hole. Using the integral-field spectrograph OSIRIS on Keck II behind the laser-guide-star adaptive optics system, this spectroscopic survey enables us to separate early-type (young, 4-6 Myr) and late-type (old, >1 Gyr) stars with a completeness of 50% down to K' = 15.5 mag, which corresponds to {approx}10 M {sub Sun} for the early-type stars. This work increases the radial extent of reported OSIRIS/Keck measurements by more than a factor of three from 4'' to 14'' (0.16 to 0.56 pc), along the projected disk of young stars. For our analysis, we implement a new method of completeness correction using a combination of star-planting simulations and Bayesian inference. We assign probabilities for the spectral type of every source detected in deep imaging down to K' = 15.5 mag using information from spectra, simulations, number counts, and the distribution of stars. The inferred radial surface-density profiles, {Sigma}(R){proportional_to}R {sup -{Gamma}}, for the young stars and late-type giants are consistent with earlier results ({Gamma}{sub early} = 0.93 {+-} 0.09, {Gamma}{sub late} = 0.16 {+-} 0.07). The late-type surface-density profile is approximately flat out to the edge of the survey. While the late-type stellar luminosity function is consistent with the Galactic bulge, the completeness-corrected luminosity function of the early-type stars has significantly more young stars at faint magnitudes compared with previous surveys with similar depth. This luminosity function indicates that the corresponding mass function of the young stars is likely less top-heavy than that inferred from previous surveys.

79 ASTRONOMY AND ASTROPHYSICS↗

A novel determination of the local dark matter density

We present a novel study on the problem of constructing mass models for the Milky Way, concentrating on features regarding the dark matter halo component. We have considered a variegated sample of dynamical observables for the Galaxy, including several results which have appeared recently, and studied a 7- or 8-dimensional parameter space - defining the Galaxy model - by implementing a Bayesian approach to the parameter estimation based on a Markov Chain Monte Carlo method. The main result of this analysis is a novel determination of the local dark matter halo density which, assuming spherical symmetry and either an Einasto or an NFW density profile is found to be around 0.39 GeV cm{sup −3} with a 1-σ error bar of about 7%; more precisely we find a ρ{sub DM}(R{sub 0}) = 0.385±0.027 GeV cm{sup −3} for the Einasto profile and ρ{sub DM}(R{sub 0}) = 0.389±0.025 GeV cm{sup −3} for the NFW. This is in contrast to the standard assumption that ρ{sub DM}(R{sub 0}) is about 0.3 GeV cm{sup −3} with an uncertainty of a factor of 2 to 3. A very precise determination of the local halo density is very important for interpreting direct dark matter detection experiments. Indeed the results we produced, together with the recent accurate determination of the local circular velocity, should be very useful to considerably narrow astrophysical uncertainties on direct dark matter detection.

79 ASTRONOMY AND ASTROPHYSICS↗

Cosmological neutrino mass: a frequentist overview in light of DESI

We derive constraints on the neutrino mass using a variety of recent cosmological datasets, including DESI BAO, the full-shape analysis of the DESI matter power spectrum and the one-dimensional power spectrum of the Lyman-α forest (P1D) from eBOSS quasars as well as the cosmic microwave background (CMB). The constraints are obtained in the frequentist formalism by constructing profile likelihoods and applying the Feldman-Cousins prescription to compute confidence intervals. This method avoids potential prior and volume effects that may arise in a comparable Bayesian analysis. Parabolic fits to the profiles allow one to distinguish changes in the upper limits from variations in the constraining power σ of the different data combinations. We find that all profiles in the ΛCDM model are cut off by the ∑m ν ≥ 0 bound, meaning that the corresponding parabolas reach their minimum in the unphysical sector. The most stringent 95% C.L. upper limit is obtained by the combination of DESI DR2 BAO, Planck PR4 and CMB lensing at 53 meV, below the minimum of 59 meV set by the normal ordering. The corresponding constraining power σ is 43 meV, which highlights the importance of the cut-off by negative values in the determination of the upper limit. Extending ΛCDM to non-zero curvature and w 0 w a CDM relaxes the constraints past 59 meV again, but only w 0 w a CDM exhibits profiles with a minimum at a positive value. Additionally, we extend the formalism to constrain the lightest neutrino mass. For DESI DR2 BAO, Planck PR4 and CMB lensing, we find confidence limits at 20 and 19 meV for normal and inverted ordering, respectively. Using a combination of DESI DR1 full-shape, BBN and eBOSS Lyman-α P1D, we successfully constrain the neutrino mass independently of the CMB. This combination yields m l ≤ 97 and 98 meV in the normal and inverted orderings, and total neutrino mass ∑m ν ≤ 285 meV (95% C.L.). The addition of DESI full-shape or Lyman-α P1D to CMB and DESI BAO results in small but noticeable improvement of the constraining power of the data. Lyman-α free-streaming measurements especially improve the constraint. Since they are based on eBOSS data, this sets a promising precedent for upcoming DESI data.

Frequentist statistics↗

Improved Bayesian Infrasonic Source Localization for regional infrasound

The Bayesian Infrasonic Source Localization (BISL) methodology is examined and simplified providing a generalized method of estimating the source location and time for an infrasonic event and the mathematical framework is used therein. The likelihood function describing an infrasonic detection used in BISL has been redefined to include the von Mises distribution developed in directional statistics and propagation-based, physically derived celerity-range and azimuth deviation models. Frameworks for constructing propagation-based celerity-range and azimuth deviation statistics are presented to demonstrate how stochastic propagation modelling methods can be used to improve the precision and accuracy of the posterior probability density function describing the source localization. Infrasonic signals recorded at a number of arrays in the western United States produced by rocket motor detonations at the Utah Test and Training Range are used to demonstrate the application of the new mathematical framework and to quantify the improvement obtained by using the stochastic propagation modelling methods. Moreover, using propagation-based priors, the spatial and temporal confidence bounds of the source decreased by more than 40 per cent in all cases and by as much as 80 per cent in one case. Further, the accuracy of the estimates remained high, keeping the ground truth within the 99 per cent confidence bounds for all cases.

58 GEOSCIENCES↗

Laplacian Smoothing Stochastic Gradient Markov Chain Monte Carlo

As an important Markov chain Monte Carlo (MCMC) method, the stochastic gradient Langevin dynamics (SGLD) algorithm has achieved great success in Bayesian learning and posterior sampling. Furthermore, SGLD typically suffers from a slow convergence rate due to its large variance caused by the stochastic gradient. In order to alleviate these drawbacks, we leverage the recently developed Laplacian smoothing technique and propose a Laplacian smoothing stochastic gradient Langevin dynamics (LS-SGLD) algorithm. We prove that for sampling from both log-concave and non-log-concave densities, LS-SGLD achieves strictly smaller discretization error in 2-Wasserstein distance, although its mixing rate can be slightly slower. Experiments on both synthetic and real datasets verify our theoretical results and demonstrate the superior performance of LS-SGLD on different machine learning tasks including posterior sampling, Bayesian logistic regression, and training Bayesian convolutional neural networks.

97 MATHEMATICS AND COMPUTING↗

Neutron multiplicity counting: Confidence intervals for reconstruction parameters

From nuclear materials accountability to homeland security, the need for improved nuclear material detection, assay, and authentication has grown over the past decades. Starting in the 1940s, neutron multiplicity counting techniques have enabled quantitative evaluation of masses and multiplications of fissile materials. In this paper, we propose a new method to compute uncertainties on these parameters using a model-based sequential Bayesian processor, resulting in credible regions in the fissile material mass and multiplication space. These uncertainties will enable us to evaluate quantitatively proposed improvements to the theoretical fission chain model. Additionally, because the processor can calculate uncertainties in real time, it is a useful tool in applications such as portal monitoring: monitoring can stop as soon as a preset confidence of non-threat is reached.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Decentralized Signal Processing and Distributed Control of Collaborative Autonomous Networks (17-ERD-101 LDRD Final Report)

Intelligent autonomous sensor networks, often comprised of large numbers of sensors, must be capable of jointly exploiting data collected at each agent in the network, and using that data to optimize their future actions towards multiple mission objectives. Centralized signal processing and optimization solutions process all data and determine all future actions at a single agent, and the resulting information and commands are disseminated back to the network. The communications bandwidth this requires and the single point of failure the central agent represents often make these solutions untenable for national security applications. In this project, several fundamental algorithms for solving both the decentralized signal processing and network optimization were developed, as well as simulation software to validate the results of these algorithms at scale. Specifically, novel algorithms for Bayesian decentralized estimation and decentralized detection and optimization based on the alternating direction method of multipliers (ADMM) were developed for autonomous sensor networks and published in the literature. The first large scale simulation of autonomous sensor networks (1000 agents) was conducted on this project, validating the performance of the developed algorithms. These algorithms and simulation tools are critical components of any decentralized autonomous network and have current and future national security applications, including distributed sensor networks for detection, estimation, and tracking problems, and large decentralized cyber-physical infrastructure such as the power grid.

42 ENGINEERING↗

DETERMINING AGES OF APOGEE GIANTS WITH KNOWN DISTANCES

We present a sample of 705 local giant stars observed using the New Mexico State University 1 m telescope with the Sloan Digital Sky Survey-III/Apache Point Observatory Galactic Evolution Experiment (APOGEE) spectrograph, for which we estimate stellar ages and the local star formation history (SFH). The high-resolution (R ∼ 22,500), near infrared (1.51–1.7 μm) APOGEE spectra provide measurements of stellar atmospheric parameters (temperature, surface gravity, [M/H], and [α/M]). Due to the smaller uncertainties in surface gravity possible with high-resolution spectra and accurate Hipparcos distance measurements, we are able to calculate the stellar masses to within 30%. For giants, the relatively rapid evolution up the red giant branch allows the age to be constrained by the mass. We examine methods of estimating age using both the mass–age relation directly and a Bayesian isochrone matching of measured parameters, assuming a constant SFH. To improve the SFH prior, we use a hierarchical modeling approach to constrain the parameters of the model SFH using the age probability distribution functions of the data. The results of an α-dependent Gaussian SFH model show a clear age–[α/M] relation at all ages. Using this SFH model as the prior for an empirical Bayesian analysis, we determine ages for individual stars. The resulting age–metallicity relation is flat, with a slight decrease in [M/H] at the oldest ages and a ∼0.5 dex spread in metallicity across most ages. For stars with ages ≲1 Gyr we find a smaller spread, consistent with radial migration having a smaller effect on these young stars than on the older stars.

79 ASTRONOMY AND ASTROPHYSICS↗

Data acquisition and path selection decision making for an autonomous roving vehicle

Problems related to the guidance of an autonomous rover for unmanned planetary exploration were investigated. Topics included in these studies were: simulation on an interactive graphics computer system of the Rapid Estimation Technique for detection of discrete obstacles; incorporation of a simultaneous Bayesian estimate of states and inputs in the Rapid Estimation Scheme; development of methods for estimating actual laser rangefinder errors and their application to date provided by Jet Propulsion Laboratory; and modification of a path selection system simulation computer code for evaluation of a hazard detection system based on laser rangefinder data.

Frederick, D. K.↗

Dislocation model for aseismic fault slip in the transverse ranges of Southern California

Geodetic data at a plate boundary can reveal the pattern of subsurface displacements that accompany plate motion. These displacements are modelled as the sum of rigid block motion and the elastic effects of frictional interaction between blocks. The frictional interactions are represented by uniform dislocation on each of several rectangular fault patches. The block velocities and fault parameters are then estimated from geodetic data. Bayesian inversion procedure employs prior estimates based on geological and seismological data. The method is applied to the Transverse Ranges, using prior geological and seismological data and geodetic data from the USGS trilateration networks. Geodetic data imply a displacement rate of about 20 mm/yr across the San Andreas Fault, while the geologic estimates exceed 30 mm/yr. The prior model and the final estimates both imply about 10 mm/yr crustal shortening normal to the trend of the San Andreas Fault. Aseismic fault motion is a major contributor to plate motion. The geodetic data can help to identify faults that are suffering rapid stress accumulation; in the Transverse Ranges those faults are the San Andreas and the Santa Susana.

Cheng, A.↗

Sparse Bayesian Regression with Integrated Feature Selection for Nuclear Reactor Analysis

High-dimensional, nonlinear function estimation using large datasets is a current area of interest in the machine learning community, and applications may be found throughout the analytical sciences, where ever-growing datasets are making more information available to the analyst. In this paper, we leverage the existing relevance vector machine, a sparse Bayesian version of the well-studied support vector machine, and expand the method to include integrated feature selection and automatic function shaping. These innovations produce an algorithm that is able to distinguish variables that are useful for making predictions of a response from variables that are unrelated or confusing. We test the technology using synthetic data, conduct initial performance studies, and develop a model capable of making position-independent predictions of the coreaveraged burnup using a single specimen drawn randomly from a nuclear reactor core.

Dayman, Ken J↗