Search NASA⌕ Search

SEARCH · Search NASA

Results for “Variational inference”

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 91 records · Page 5

Variations of the stellar initial mass function in the progenitors of massive early-type galaxies and in extreme starburst environments

We examine variations of the stellar initial mass function (IMF) in extreme environments within the formalism derived by Hennebelle and Chabrier. We focus on conditions encountered in progenitors of massive early-type galaxies and starburst regions. We show that, when applying the concept of turbulent Jeans mass as the characteristic mass for fragmentation in a turbulent medium, the peak of the IMF in such environments is shifted toward smaller masses, leading to a bottom-heavy IMF, as suggested by various observations. In very dense and turbulent environments, we predict that the high-mass tail of the IMF can become even steeper than the standard Salpeter IMF, with a limit for the power-law exponent α ≅ –2.7, in agreement with recent observational determinations. This steepening is a direct consequence of the high densities and Mach values in such regions but also of the time dependence of the fragmentation process, as incorporated in the Hennebelle-Chabrier theory. We provide analytical parameterizations of these IMFs in such environments to be used in galaxy evolution calculations. We also calculate the star-formation rates and the mass-to-light ratios expected under such extreme conditions and show that they agree well with the values inferred in starburst environments and massive high-redshift galaxies. This reinforces the paradigm of star formation as being a universal process, i.e., the direct outcome of gravitationally unstable fluctuations in a density field initially generated by large-scale, shock-dominated turbulence. This globally enables us to infer the variations of the stellar IMF and related properties for atypical galactic conditions.

79 ASTRONOMY AND ASTROPHYSICS↗

Adversarial uncertainty quantification in physics-informed neural networks

Here, we introduce a deep learning framework for quantifying and propagating uncertainty in systems governed by non-linear differential equations using physics-informed neural networks. Specifically, we employ latent variable models to construct probabilistic representations for the system states, and put forth an adversarial inference procedure for training them on data, while constraining their predictions to satisfy given physical laws expressed by partial differential equations. Such physics-informed constraints provide a regularization mechanism for effectively training deep generative models as surrogates of physical systems in which the cost of data acquisition is high, and training data-sets are typically small. This provides a flexible framework for characterizing uncertainty in the outputs of physical systems due to randomness in their inputs or noise in their observations that entirely bypasses the need for repeatedly sampling expensive experiments or numerical simulators. Moreover, we demonstrate the effectiveness of our approach through a series of examples involving uncertainty propagation in non-linear conservation laws, and the discovery of constitutive laws for flow through porous media directly from noisy data.

97 MATHEMATICS AND COMPUTING↗

Improving Seasonal Forecast Using Probabilistic Deep Learning

The path toward realizing the potential of seasonal forecasting and its socioeconomic benefits relies on improving general circulation model (GCM) based dynamical forecast systems. To improve dynamical seasonal forecasts, it is crucial to set up forecast benchmarks, and clarify forecast limitations posed by model initialization errors, formulation deficiencies, and internal climate variability. With huge costs in generating large forecast ensembles, and limited observations for forecast verification, the seasonal forecast benchmarking and diagnosing task proves challenging. Here, we develop a probabilistic deep learning-based statistical forecast methodology, drawing on a wealth of climate simulations to enhance seasonal forecast capability and forecast diagnosis. By explicitly modeling the internal climate variability and GCM formulation differences, the proposed Conditional Generative Forecasting (CGF) methodology enables bypassing crucial barriers in dynamical forecast, and offers a top-down viewpoint to examine how complicated GCMs encode the seasonal predictability information. We apply the CGF methodology for global seasonal forecast of precipitation and 2 m air temperature, based on a unique data set consisting 52,201 years of climate simulation. Results show that the CGF methodology can faithfully represent the seasonal predictability information encoded in GCMs. We successfully apply this learned relationship in real-world seasonal forecast, achieving competitive performance compared to dynamical forecasts. Using this CGF as benchmark, we reveal the impact of insufficient forecast spread sampling that limits the skill of the considered dynamical forecast system. Finally, we introduce different strategies for composing ensembles using the CGF methodology, highlighting the potential for leveraging the strengths of multiple GCMs to achieve advantgeous seasonal forecast.

54 ENVIRONMENTAL SCIENCES↗

Distributed Acoustic Sensing to Estimate the Permeability

Optical fiber in a borehole can be interrogated with distributed acoustic sensors (DAS) to capture fracture displacements with the potential to map surrounding fracture networks. We designed a laboratory experiment to test the capability of DAS to determine borehole flow characteristics, and we show that for the first time DAS can be used to remotely estimate permeability. Optical fiber was wrapped around a bead filled pipe and the pressure drop and flow velocity were measured to directly calculate permeability. A machine learning model using statistical features from continuous DAS estimated the bulk permeability. Fluid interactions with the permeable material demonstrate insufficient resolution using DAS amplitude-based measurements for estimating pressure drop to infer permeability. Variations in the spectral domain relate DAS measurements to the pressure drop and provide consistent permeability estimates. Resolution with DAS is sufficient to estimate permeability and provides a reliable method to monitor at depth in borehole conditions.

58 GEOSCIENCES↗

High speed measurements of neutral beam turn-on and impact of beam modulation on measurements of ion density

Modulation of neutral beams on tokamaks is performed routinely, enabling background rejection for active spectroscopic diagnostics, and control of injected power and torque. We find that there exists an anomalous initial transient in the beam neutrals delivered to the tokamak that is not accounted for by the accelerator voltage and power supply current. Measurements of the charge-exchange and beam photoemission on the DIII-D tokamak [J. L. Luxon, Nucl. Fusion 42, 614 (2002)] at high speed (200 μs) reveal that the energy of the beam neutrals is constant, but the density of beam neutrals displays dramatic variation in the first 2–3 ms following beam turn-on. The impact of this beam density variation on inferred ion densities and impurity transport is presented, with suggested means to correct for the anomalous transient.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

A convergence metric for counting statistics in time-resolved small angle neutron scattering

Here, this work introduces a model-independent, dimensionless metric for predicting optimal measurement duration in time-resolved small-angle neutron scattering using early-time data. Built on a Gaussian process regression framework, the method reconstructs scattering profiles with quantified uncertainty, even from sparse or noisy measurements. Demonstrated on the EQ-SANS instrument at the Spallation Neutron Source, the approach generalizes to general SANS instruments with a two-dimensional detector. A key result is the discovery of a dimensionless convergence metric revealing a universal power-law scaling in profile evolution across soft matter systems. When time is normalized by a system-specific characteristic time t*, the variation in inferred profiles collapses onto a single curve with an exponent between −2 and −1. This trend emerges within the first ten time steps, enabling early prediction of measurement sufficiency. The method supports real-time experimental optimization and is especially valuable for maximizing efficiency in low-flux environments such as compact accelerator-based neutron sources.

Tung, Chi-Huan [Oak Ridge National Laboratory (ORN↗

EXTREME CORONAL LINE EMITTERS: TIDAL DISRUPTION OF STARS BY MASSIVE BLACK HOLES IN GALACTIC NUCLEI?

Tidal disruption of stars by supermassive black holes at the centers of galaxies is expected to produce unique emission-line signatures, which have not yet been explored adequately. Here we report the discovery of extremely strong coronal lines from [Fe X] up to [Fe XIV] in a sample of seven galaxies (including two recently reported cases), which we interpret as such signatures. This is the first systematic search for objects of this kind, by making use of the immense database of the Sloan Digital Sky Survey. The galaxies, which are non-active as evidenced by the narrow line ratios, show broad emission lines of complex profiles in more than half of the sample. Both the high-ionization coronal lines and the broad lines are fading on timescales of years in objects observed with spectroscopic follow-ups, suggesting their transient nature. Variations of inferred non-stellar continua, which have absolute magnitudes of at least -16 to -18 mag in the g band, are also detected in more than half of the sample. The coronal line emitters reside in sub-L{sub *} disk galaxies (-21.3 < M{sub i} < -18.5) with small stellar velocity dispersions. The sample seems to form two distinct types based on the presence or absence of the [Fe VII] lines, with the latter having relatively low luminosities of [O III], [Fe XI], and the host galaxies. These characteristics can most naturally be understood in the context of transient accretion onto intermediate-mass black holes at galactic centers following tidal disruption of stars in a gas-rich environment. We estimate the incidence of such events to be around 10{sup -5} yr{sup -1} for a galaxy with -21.3 < M{sub i} < -18.5.

79 ASTRONOMY AND ASTROPHYSICS↗

LARGER PLANET RADII INFERRED FROM STELLAR ''FLICKER'' BRIGHTNESS VARIATIONS OF BRIGHT PLANET-HOST STARS

Most extrasolar planets have been detected by their influence on their parent star, typically either gravitationally (the Doppler method) or by the small dip in brightness as the planet blocks a portion of the star (the transit method). Therefore, the accuracy with which we know the masses and radii of extrasolar planets depends directly on how well we know those of the stars, the latter usually determined from the measured stellar surface gravity, log g. Recent work has demonstrated that the short-timescale brightness variations ({sup f}licker{sup )} of stars can be used to measure log g to a high accuracy of ∼0.1-0.2 dex. Here, we use flicker measurements of 289 bright (Kepmag < 13) candidate planet-hosting stars with T {sub eff} = 4500-6650 K to re-assess the stellar parameters and determine the resulting impact on derived planet properties. This re-assessment reveals that for the brightest planet-host stars, Malmquist bias contaminates the stellar sample with evolved stars: nearly 50% of the bright planet-host stars are subgiants. As a result, the stellar radii, and hence the radii of the planets orbiting these stars, are on average 20%-30% larger than previous measurements had suggested.

79 ASTRONOMY AND ASTROPHYSICS↗

Estimating fluid-induced stress change from observed deformation

Observed deformation is sensitive to a changing stress field within the Earth. There are, however, several impediments to a direct inversion of geodetic measurements for changes in stress. Estimating six independent components of stress change from a smaller number of displacement or strain components is inherently non-unique. The reliance upon surface measurements leads to a loss of resolution, due to the attenuation of higher spatial frequencies in the displacement field with distance from a source. Here, we adopt a technique suited to the estimation of stress changes due to the injection and/or withdrawal of fluids at depth. In this approach the surface displacement data provides an estimate of the volume change responsible for the deformation, rather than stress changes themselves. The inversion for volume change is constrained by the fluid fluxes into and out of the reservoir. The distribution of volume change is used to calculate the displacements in the region above the reservoir. Estimates of stress change follow from differentiating the displacement field in conjunction with a geomechanical model of the o verburden. We also apply the technique to Interferometric Synthetic Aperture Radar (InSAR) observations gathered over a petroleum reservoir in the San Joaquin Valley of California. An analysis of the InSAR range changes reveals that the stress field in the overburden varies rapidly both in space and in time. The inferred stress variations are found to be compatible with the documented failure of a well in the field.

58 GEOSCIENCES↗

Posterior Covariance Matrix Approximations

Here, the Davis equation of state (EOS) is commonly used to model thermodynamic relationships for high explosive (HE) reactants. Typically, the parameters in the EOS are calibrated, with uncertainty, using a Bayesian framework and Markov Chain Monte Carlo (MCMC) methods. However, MCMC methods are computationally expensive, especially for complex models with many parameters. This paper provides a comparison between MCMC and less computationally expensive Variational methods (Variational Bayesian and Hessian Variational Bayesian) for computing the posterior distribution and approximating the posterior covariance matrix based on heterogeneous experimental data. All three methods recover similar posterior distributions and posterior covariance matrices. This study demonstrates that for this EOS parameter calibration application, the assumptions made in the two Variational methods significantly reduce the computational cost but do not substantially change the results compared to MCMC.

97 MATHEMATICS AND COMPUTING↗

CONCURRENT, CONDENSED STEIN VARIATIONAL GRADIENT DESCENT FOR UNCERTAINTY QUANTIFICATION OF NEURAL NETWORKS

In this work, we propose a Stein variational gradient descent (SVGD) method to concurrently sparsify, train, and provide uncertainty quantification (UQ) of a complexly parameterized model, such as a neural network (NN). It employs a graph reconciliation and condensation process to reduce complexity and increase similarity in the Stein ensemble of parameterizations. Therefore, the proposed concurrent, condensed SVGD (ccSVGD) method can provide UQ on parameters, not just outputs. Furthermore, the parameter reduction speeds up the convergence of the Stein gradient descent as it reduces the combinatorial complexity by aligning and differentiating the sensitivity to parameters. These properties are demonstrated with an illustrative example and an application to a mechanical response representation problem in solid mechanics.

42 ENGINEERING↗

We Know Where We Don't Know: 3D Bayesian CNNs for Credible Geometric Uncertainty

Deep learning has been successfully applied to the segmentation of 3D Computed Tomography (CT) scans. Establishing the credibility of these segmentations requires uncertainty quantification (UQ) to identify untrustworthy predictions. Recent UQ architectures include Monte Carlo dropout networks (MCDNs), which approximate deep Gaussian processes, and Bayesian neural networks (BNNs), which learn the distribution of the weight space. BNNs are advantageous over MCDNs for UQ but are thought to be computationally infeasible in high dimension, and neither architecture has produced interpretable geometric uncertainty maps. We propose a novel 3D Bayesian convolutional neural network (BCNN), the first deep learning method which generates statistically credible geometric uncertainty maps and scales for application to 3D data. We present experimental results on CT scans of graphite electrodes and laser-welded metals and show that our BCNN outperforms an MCDN in recent uncertainty metrics. The geometric uncertainty maps generated by our BCNN capture distributions of sigmoid values that are interpretable as confidence intervals, critical for applications that rely on deep learning for high-consequence decisions.

97 MATHEMATICS AND COMPUTING↗

Xenon in the Protoplanetary Disk (PPD), in Two Planets, and a Comet

Isotopic anomalies in several elements, as recently observed in meteorites, are generally interpreted to indicate nonequilibrium environments in the protoplanetary disk (PPD). Here we study reported Xe isotopic compositions on planets Earth and Mars, in a comet, and in meteorites for precursor discrepancies. Abundance variations of inferred presolar nano-diamonds, the carrier phase of the Xe-HL component, appear to be the primary source of nonuniformity of Xe precursors in the PPD, together with mechanisms of mass-dependent fractionation. While planet Mars kept a record of initial solar Xe isotopic abundances, such a record is missing for planet Earth. Xe isotopic abundances in paleo-atmospheres of both planets represent secondary reservoirs that show mass-dependent fractionation effects, but the inferred compositions of their PPD precursors differ: Mars atmospheric precursor Xe had solar isotopic composition, while Earth’s Xe precursor is consistent with a PPD reservoir of low nano-diamond abundance. Strong mass-dependent fractionation effects are observed in Xe components of IAB irons and in Yamato carbonaceous (CY) chondrites, and show that fractionation mechanisms are not restricted to planetary atmospheres. These records show that Xe isotopes in solar system reservoirs are useful tracers of evolutionary processes and of nonequilibrated presolar components in the PPD.

79 ASTRONOMY AND ASTROPHYSICS↗

SQUID Readout of a High-$Q$ Superconducting $LC$ Resonator

We demonstrate the readout of a superconducting $LC$ resonator, with $Q \approx 2 \times 10^{6}$ at approximately 250 kHz, using a dc SQUID followed by a SQUID series-array amplifier readout chain. We find that $Q$ depends on the SQUID flux-bias point, increasing near the shallow-slope point of the SQUID modulation curve and decreasing near the steep-slope point, consistent with the previously observed SQUID damping effects. From this variation, we infer the SQUID effective input impedance. We further infer the effective temperature of the resonator circuit from the resonance peak in the noise spectrum and show that it also depends on the SQUID flux-bias point, suggesting a possible contribution from SQUID back-action noise. This system provides a working prototype for future experiments based on lumped-element resonators with SQUID readout, in particular low-mass axion searches within the DMRadio program.

FOS: Physical sciences↗

Subsurface Seismic Structure Around the Sanford Underground Research Facility

To address the fundamental challenges of understanding processes associated with enhanced or engineered geothermal systems (EGS), teams from the EGS Collab SIGMA-V project are conducting stimulations at the Sanford Underground Research Facility. Various modalities of geophysical observations are collected and analyzed, and different types of parameters are modeled to help investigate these critical processes. Among these parameters, the three-dimensional stress field is one of the primary objectives. Seismic velocity models have been used to estimate the 3D stress field for other regions. A detailed seismic velocities and density model can help us infer spatial variations of the stress field in the study area. We inverted the 3D seismic structure around the site by using body-wave travel times, surface-wave dispersions and gravity observations simultaneously. The resulting model will be a starting point for more detailed investigations and could also be used to study subsurface geological and stress field variations of the broad area.

Chai, Chengping↗

Subsurface Seismic Structure Around the Sanford Underground Research Facility

To address the fundamental challenges of understanding processes associated with enhanced or engineered geothermal systems (EGS), teams from the EGS Collab SIGMA-V project are conducting stimulations at the Sanford Underground Research Facility. Various modalities of geophysical observations are collected and analyzed, and different types of parameters are modeled to help investigate these critical processes. Among these parameters, the three-dimensional stress field is one of the primary objectives. Seismic velocity models have been used to estimate the 3D stress field for other regions. A detailed seismic velocities and density model can help us infer spatial variations of the stress field in the study area. We inverted the 3D seismic structure around the site by using body-wave travel times, surface-wave dispersions and gravity observations simultaneously. The resulting model will be a starting point for more detailed investigations and could also be used to study subsurface geological and stress field variations of the broad area.

Chai, Chengping↗