Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “gradient sampling”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
NASA NTRS2002

Overview of ACE-Asia Spring 2001 Investigations On Aerosol-Radiation Interactions

ACE-Asia's extensive measurements from land, ocean, air and space quantified aerosol-radiation interactions. Results from each platform type, plus satellite-suborbital combinations, include: 1. Time series of multiwavelength aerosol optical depth (ADD), Angstrom exponent (alpha), single-scattering albedo (SSA), and size distribution from AERONET radiometry at 13 stations. In China and Korea AOD and alpha were strongly anticorrelated (reflecting transient dust events); dust volume-size modes peaked near 8 microns diameter; and SSA(dust) greater than SSA(pollution). 2. Calculations and measurements of photosynthetically active radiation and aerosols in China yield 24-h average downward surface radiative forcing per AOD(500 nm) of -27 W/sq m (400-700 nm). 3. The Hawaii-Japan cruise sampled a gradient with AOD(500 nm) extremes of 0.1 and 1.1. Shipboard measurements showed that adding dust to pollution increased SSA(550 nm, 55% RH), typically from -0.91 to approx. 0.97. Downwelling 8-12 micron radiances showed aerosol effects, especially in the major April dust event, with longwave forcing estimated at -5 to 15 W/sq m. 4. Extinction profiles from airborne sunphotometry and total-direct-diffuse radiometry show wavelength dependence often varying strongly with height, reflecting layering of dust-dominated over pollution-dominated aerosols. Comparing sunphotometric extinction profiles to those from in situ measurements (number and composition vs size, or scattering and absorption) shows layer heights agree, but extinction sometimes differs. 5. Airborne solar spectral flux radiometry yields absorption spectra for layers. Combining with AOD spectra yields best-fit aerosol single scattering albedo spectra. 6. Visible, NIR and total solar fluxes combined with AOD give radiative forcing efficiencies at surface and aloft.

Russell, P. B.↗
DOE OSTI2022

Colour gradients of low-redshift galaxies in the DESI Legacy Imaging Survey

ABSTRACT Radial colour gradients within galaxies arise from gradients of stellar age, metallicity, and dust reddening. Large samples of colour gradients from wide-area imaging surveys can complement smaller integral-field spectroscopy data sets and can be used to constrain galaxy formation models. Here, we measure colour gradients for low-redshift galaxies (z < 0.1) using photometry from the DESI Legacy Imaging Survey DR9. Our sample comprises ∼93 000 galaxies with spectroscopic redshifts and ∼574 000 galaxies with photometric redshifts. We focus on gradients across a radial range 0.5Reff to Reff, which corresponds to the inner disc of typical late-type systems at low redshift. This region has been the focus of previous statistical studies of colour gradients and has recently been explored by spectroscopic surveys such as MaNGA. We find that the colour gradients of most galaxies in our sample are negative (redder towards the centre), consistent with the literature. We investigate empirical relationships between colour gradient, average g − r and r − z colour, Mr, M⋆, and sSFR. Trends of gradient strength with Mr (M⋆) show an inflection around Mr ∼ −21 ($\log _{10} \, M_\star /\mathrm{M_\odot }\sim 10.5$). Below this mass, colour gradients become steeper with increasing M⋆, whereas colour gradients in more massive galaxies become shallower. We find that positive gradients (bluer stars at smaller radii) are typical for galaxies of $M_{\star }\sim 10^{8}\, \mathrm{M_\odot }$. We compare our results to age and metallicity gradients in two data sets derived from fits of different stellar population libraries to MaNGA spectra, but find no clear consensus explanation for the trends we observe. Both MaNGA data sets seem to imply a significant contribution from dust reddening, in particular, to explain the flatness of colour gradients along the red sequence.

Astronomy & Astrophysics↗
DOE OSTI2024

Variance-Reduced Accelerated First-Order Methods: Central Limit Theorems and Confidence Statements

In this paper, we consider a strongly convex stochastic optimization problem and propose three classes of variable sample-size stochastic first-order methods: (i) the standard stochastic gradient descent method, (ii) its accelerated variant, and (iii) the stochastic heavy-ball method. In each scheme, the exact gradients are approximated by averaging across an increasing batch size of sampled gradients. We prove that when the sample size increases at a geometric rate, the generated estimates converge in mean to the optimal solution at an analogous geometric rate for schemes (i)–(iii). Based on this result, we provide central limit statements, whereby it is shown that the rescaled estimation errors converge in distribution to a normal distribution with the associated covariance matrix dependent on the Hessian matrix, the covariance of the gradient noise, and the step length. If the sample size increases at a polynomial rate, we show that the estimation errors decay at a corresponding polynomial rate and establish the associated central limit theorems (CLTs). Under certain conditions, we discuss how both the algorithms and the associated limit theorems may be extended to constrained and nonsmooth regimes. As a result, we provide an avenue to construct confidence regions for the optimal solution based on the established CLTs and test the theoretical findings on a stochastic parameter estimation problem.

Lei, Jinlong↗
DOE OSTI2026

A dual dynamic shutter system for accelerating ion irradiation sample throughput via lateral gas implantation gradients

Ion irradiation for material performance testing is limited due to its serial nature, which allows for only one value of the implantation (appm) versus dose (dpa) parameter space to be explored for each ion and experiment at a time. While ion irradiation can accelerate the process by up to three orders of magnitude compared to neutron irradiation experiments, the sample throughput for ion irradiation remains relatively low. To address these limitations, a novel capability has been developed at the Michigan Ion Beam Laboratory (MIBL), enabling for the creation of single- and two-dimensional lateral ion implantation gradients using recently installed motorized-controlled ion-beam shutters. This advancement can generate a wide scope of the two-dimensional (H+, He2+) implantation parameter space within a single sample. Integration of this new capability now allows for dual- and triple-ion beam experiments to be performed with full user control over not only the ion implantation depth, but also laterally across the sample by imposing ion implantation concentration gradients, thus providing researchers with a high-throughput means for material testing under various irradiation conditions. Furthermore, recent improvements in MIBL's microbeam ion-beam analysis (IBA) target station now allow for probing these concentration gradients in irradiated alloys with exceptional spatial resolution, down to 10 µm. These two approaches promise to significantly improve ion irradiation capabilities and increase the sample throughput by several orders of magnitude. The application of the shutter technique plus the subsequent microbeam characterization of the imposed implantation gradients are showcased by two proof-of-principle ion-irradiated experiments, one performed on single-crystal Si and the other on the fusion-candidate alloy F82H-IEA. These advancements mark a substantial leap in ion-beam technology, offering researchers a robust, high-throughput method to efficiently investigate candidate alloys with high technological readiness for both advanced fission and fusion reactor applications, in a time- and cost-effective manner.

36 - MATERIALS SCIENCE↗
NASA NTRS1984

Directional solidification of flake and nodular cast iron during KC-135 low-g maneuvers

Alloys solidified in a low-gravity environment can, due to the elimination of sedimentation and convection, form unique and often desirable microstructures. One method of studying the effects of low-gravity (low-g) on alloy solidification was the use of the NASA KC-135 aircraft flying repetitive low-g maneuvers. Each maneuver gives from 20 to 30 seconds of low-g which is between about 0.1 and 0.001 gravity. A directional solidification furnace was used to study the behavior of off eutectic composition case irons in a low-g environment. The solidification interface of hypereutectic flake and spheroidal graphite case irons was slowly advanced through a rod sample, 5 mm in diameter. Controlled solidification was continued through a number of aircraft parabolas. The known solidification rate of the sample was then correlated with accelerometer data to determine the gravity level during solidification for any location of the sample. The thermal gradient and solidification rate were controlled independently. Samples run on the KC-135 aircraft exhibited bands of coarser graphite or of larger nodules usually corresponding to the regions solidified under low-g. Samples containing high phosphorous (used in order to determine the eutectic cell) exhibited larger eutectic cells in the low-g zone, followed by a band of coarser graphite.

Curreri, P. A.↗
NASA NTRS2021

CLEAR: The Gas-phase Metallicity Gradients of Star-forming Galaxies at 0.6<z<2.6

We report on the gas-phase metallicity gradients of a sample of 238 star-forming galaxies at 0.6 < z < 2.6, measured through deep near-infrared Hubble Space Telescope slitless spectroscopy. The observations include 12 orbit depth Hubble/WFC3 G102 grism spectra taken as a part of the CANDELS Lyα Emission at Reionization (CLEAR) survey, and archival WFC3 G102+G141 grism spectra overlapping the CLEAR footprint. The majority of galaxies in this sample are consistent with having a zero or slightly positive metallicity gradient ( dZ/dR ≥ 0, i.e., increasing with radius) across the full mass range probed (8.5 < log M * / M ⊙ < 10.5). We measure the intrinsic population scatter of the metallicity gradients, and show that it increases with decreasing stellar mass—consistent with previous reports in the literature, but confirmed here with a much larger sample. To understand the physical mechanisms governing this scatter, we search for correlations between the observed gradient and various stellar population properties at fixed mass. However, we find no evidence for a correlation with the galaxy properties we consider—including star formation rates, sizes, star formation rate surface densities, and star formation rates per gravitational potential energy. We use the observed weakness of these correlations to provide material constraints for predicted intrinsic correlations from theoretical models.

Raymond C. Simons↗
NASA NTRS1985

Electric Field Containerless Processing Technology

The objective of this task is to develop the science and technology base required to design and construct a high temperature electric field positioning module that could be used by materials scientists to conduct containerless science experiments in a low gravity environment. Containerless science modules that employ electric fields to position and manipulate samples offer several advantages over acoustic or electromagnetic systems. The electric field system will operate not only at atmospheric pressures but also in a vacuum, in contrast to the acoustic modules which can only operate in atmosphere where the acoustic forces are sufficient. The electric field technique puts minimum energy into the sample, whereas the electromagnetic system can deposit energy into the sample through eddy current heat as well as physical mixing in the sample. Two types of electric field modules have been constructed and tested to date. One employs a charged sample and uses electrostatic forces to position and control the sample. The second type of module induces electrical polarization of the sample and electric field gradients to position and control the sample.

Elleman, D. D.↗
NASA NTRS1974

The use of the ion probe mass spectrometer in the measurement of hydrogen concentration gradients in Monel K 500

The ion probe mass spectrometer was used to measure hydrogen concentration gradients in cathodically charged Monel K 500. Initial work with the ion probe involved the calibration of the instrument and the establishment of a suitable experimental procedure for this application. Samples of Monel K 500 were cathodically charged in a weak sulfuric acid solution. By varying the current density, different levels of hydrogen were introduced into the samples. Hydrogen concentration gradients were taken by ion sputtering on the surface of these samples and monitoring the behavior of the hydrogen mass peak as a function of time. An attempt was made to determine the relative amounts of hydrogen in the bulk and grain boundaries by analyzing a fresh fracture surface with a higher proportion of grain boundary area. It was found that substantially more hydrogen was detected in the grain boundaries than in the bulk, confirming the predictions of previous workers. A sputter rate determination was made in order to establish the rate of erosion.

Truhan, J. J., Jr.↗
NASA NTRS1994

Thermal-infrared remote sensing and Kirchhoff's Law. 1: Laboratory measurements

By the end of this century the Earth Observing System (EOS) will provide worldwide, thermal infrared, multispectral images of the Earth, presenting geologists with a new kind of remote sensing data for interpretation. Thus it has become essential to understand the spectral emittance behavior of terrestrial surface materials. Perhaps the most fundamental question to be answereed is the extent to which such materials follow Kirchhoff's law (epsilon = 1 -R) under laboratory and field conditions, especially when a sample displays a thermal gradient. We present the first rigorous quantitative comparison of directional and hemispherical reflectance and directional emittance of rock and soil samples in the laboratory, with thermal gradients induced by heating them from below and allowing them to radiate to a colder background. The results show that only an extemeley low density sample composed of fine particles sifted into a 'fairy castle' structure displays a thermal gradient steep enough within the infrared skin depth to cause significant (6%) departure from Kirchhoff's law. There is no detectable effect on the more normal terrestrial samples, such as soils and rocks measured in the laboratory, even when semitransparent coatings are involved. Thus both emittance and reflectance measurements can be used to calculate sample emissivity for most terrestrial surface materieals. However, the effect on Kirchhoffian behavior of different field environments, which may induce a steeper thermal gradient in particulate samples, has yet to be determined, and some low-density surface materials like newly fallen snow, frost, and efflorescent salts on playas have yet to be measured in emittance.

Salisbury, John W.↗
DOE OSTI2022

Quantifying electron temperature distributions from time-integrated x-ray emission spectra

K-shell x-ray emission spectroscopy is a standard tool used to diagnose the plasma conditions created in high-energy-density physics experiments. In the simplest approach, the emissivity-weighted average temperature of the plasma can be extracted by fitting an emission spectrum to a single temperature condition. It is known, however, that a range of plasma conditions can contribute to the measured spectra due to a combination of the evolution of the sample and spatial gradients. In this work, we define a parameterized model of the temperature distribution and use Markov Chain Monte Carlo sampling of the input parameters, yielding uncertainties in the fit parameters to assess the uniqueness of the inferred temperature distribution. Here we present the analysis of time-integrated S and Fe x-ray spectroscopic data from the Orion laser facility and demonstrate that while fitting each spectral region to a single temperature yields two different temperatures, both spectra can be fit simultaneously with a single temperature distribution. We find that fitting both spectral regions together requires a maximum temperature of $1310^{+90}_{-70}$ eV with significant contributions from temperatures down to 200 eV.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2023

Determination of the hydrogen heat of transport in Zircaloy-4

During operation in a nuclear reactor, Zr-based nuclear fuel cladding is subject to waterside corrosion which can lead to hydrogen ingress. Here, the hydrogen that enters the material will migrate to colder spots and precipitate as zirconium hydrides if the hydrogen content exceeds the hydrogen terminal solid solubility in the material. Since a temperature gradient is established in the radial direction of the cladding during operation, the hydrides can preferentially precipitate at the colder outer surface of the cladding. Other gradients can also occur in the longitudinal and azimuthal directions of the cladding tube. As a consequence, hydrogen redistributes itself in response to the concentration and temperature gradients present in the sample. The response of the hydrogen in solid solution to temperature gradients is governed by the heat of transport Q* as a function of temperature, so it can be used in the BISON code. A set of experiments was set up to determine the heat of transport (Q*), in which a uniformly hydrided Zircaloy-4 sample is annealed under a fixed temperature gradient at a range of temperatures, and the resulting hydrogen distribution is analyzed to determine Q*. The results are discussed in terms of existing literature.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
NASA NTRS1999

Characterization of the Heat Extraction Capability of a Compliant, Sliding, Thermal Interface for Use in a High Temperature, Vacuum Microgravity Furnace

Much of the material science gained in microgravity research requires processing a sample in a high performance furnace. One such furnace currently being designed is the Quench Module Insert (QMI). The Principle Investigators, for whom the furnace is proposed, require high temperature gradients in their cylindrical samples to achieve the science objectives. One of the components critical to achieving high sample axial temperature gradients in the Quench Module Insert is a high performance cold zone to extract the heat from the sample. This cold zone employs a compliant, sliding thermal interface based on a Vel-Therm felt. This felt provides a conductive path between the Sample Cartridge Assembly (SCA) exterior surface and the interior surface of the water cooled chill block while allowing movement of the sample relative to the chill block. The Vel-Therm felt is composed of long polymer-based fibers affixed to a thin flexible substrate layer. The fibers are oriented perpendicular to this substrate giving the felt the appearance of a velvet fabric. The Vel-Therm felt heat extraction capability was quantified in earlier tests performed in an inert gas environment. The current activity, described in this paper, is intended to characterize the extraction capability of Vel-Therm felt in a vacuum environment similar to the QMI environment. This testing is necessary to quantify the thermal performance of the Vel-Therm felt and the sensitivity of that performance to key variables. The data derived from these tests will be incorporated into the current thermal models to improve the quality of the models and reduce uncertainty of the analytical results. In addition, the data will be used to help select the appropriate Vel-Therm felt and set proper operating limits as well as assess the performance range of the furnace. The objective of this test is to measure the heat extraction rate of the Vel-Therm felt as specified by the effective heat transfer coefficient. Therefore, the test setup was designed to force the bulk of the heat transfer through the area where the Vel-Therm felt was applied. A heat source, consisting of a ceramic heating element encased in a Copper (Cu) housing is mounted on four isolated support rods. A 6-layer molybdenum radiation shield is used to insulate against heat loss from the heater and prevent heat exchange between the hot and cold sides of the test apparatus. The Vel-Therm felt is affixed to the surface of the cold sink, a water-cooled Cu chill block. An adjustable plate supports and isolates the cold sink from above and is used to control the amount of the deflection of the Vel- Therm when in contact with the Cu heating element housing. The primary means of establishing the power being conducted through the felt is to measure the energy being transferred to the water passing through the chill block. Analysis was performed to support the assumption that the source and sink surfaces were approximately isothermal under the specific test conditions. As a check on the amount of power passing through the felt, the power supplied to the heater was also measured. Thermocouples were strategically located throughout the test apparatus for measurement purposes. A bell jar was lowered over the assembly to impose vacuum conditions. Currently, variables tested have been fiber compression and fiber type and surface temperatures (both the hot and cold side temperatures are hypothesized to be important to the performance of the Vel-Therm.) Selected runs were repeated to ensure consistency and repeatability. Results obtained thus far reveal that Vel-Therm performance is significantly degraded by fibers being exposed to high compression. It also shows that performance is somewhat negatively impacted by previous compression, thereby, raising the question of repeatability. In addition, early results show a significant dependence on temperature. A computer aided mathematical analysis of the test setup is ongoing. The results will be correlated to actual results. The correlation will examine such details as parasitic loses, conduction down the power leads and many other concerns.

Bellomy-Ezell, Jenny↗
DOE OSTI2025

Discovery, Design, Synthesis and Testing of High Performance Structural Alloys (Final Technical Report)

The overarching goal of this project is to understand the phase stability and mechanical behavior of non-stoichiometric multi-principal element alloy (MPEA) materials. In order to identify suitable alloys, we plan to use a combinatorial thin film screening approach, in collaboration with scientists at Lawrence Berkeley National Laboratory who are performing computational work as well as complementary experimental work. Specific tasks within the scope of this project include the fabrication, using thin film deposition from six sputtering targets, of combinatorial samples with multi-dimensional gradients in composition and microstructure. These samples are studied to screen MPEA systems for promising candidate alloys with specific composition(s), based on characterization of composition, structure and mechanical behavior across the thin film. We want to produce single-phase MPEAs with chemical homogeneity in a given thin film region, simple grain structures, and no intermetallic phases present. Gradient films facilitate first-pass screening for desirable characteristics and inform the next stage of work that involves fabrication of bulk MPEA specimens for (tensile) mechanical testing and characterization. To make the bulk alloys, metal (elemental) pieces are melted to form MPEAs, followed by heat treatment to homogenize the composition and microstructure. A subset of alloys is also cast, using vacuum arc melting, to yield larger samples (diameter ~1 cm and length ~5-10 cm) and these allow us to assess viability of scale-up for the alloys in structural applications. Further processing plans include rolling and heat treatment to recrystallize selected bulk MPEAs and grow grains to different extents, in order to investigate size effects in the mechanical behavior of MPEAs. Microspecimen testing will be performed (primarily in tension) to assess the mechanical behavior over a range of temperatures. The deformation microstructure of mechanically tested alloys will be characterized using transmission electron microscopy (TEM) to provide a scientific basis for understanding the structure-property relationships in MPEA mechanical behavior.

36 MATERIALS SCIENCE↗
DOE OSTI2024

A Method for Dynamic Kolsky Bar Compression at High Temperatures: Application to Ti-6Al-4V

Abstract An experimental apparatus for measuring the dynamic behavior of materials subjected to strain rates on the order of 10 $$^3$$ 3 s $$^{-1}$$ - 1 and temperatures up to 800°C with a unique triple actuation system is developed in this work. This system is based on the traditional Kolsky (or split-Hopkinson pressure) bar design, with the addition of an external furnace used to heat the specimen to the desired temperature. A synchronized triple pneumatic actuation system is used to control the motion and timing of the the sample, incident, and transmitted bars. The cold contact time (CCT), or the time during which the heated sample is in contact with the room temperature bars before compression, is measured experimentally and carefully controlled to minimize the development of a temperature gradient across the sample and avoid heating of the bars. Experiments are performed in conjunction with ultra high speed imaging and 2D digital image correlation (DIC), as well as high speed thermal imaging. To verify the viability of the proposed system, experiments were conducted on Ti-6Al-4V (wt.%) at temperatures from 25°C up to and 800°C at an average strain rate of approximately 1200 s $$^{-1}$$ - 1 .

36 MATERIALS SCIENCE↗
NASA NTRS2007

Early Results and Spaceflight Implications of the SWAB Flight Experiment

Microbial monitoring of spacecraft environments provides key information in the assessment of infectious disease risk to the crew. Monitoring aboard the Mir space station and International Space Station (ISS) has provided a tremendous informational baseline to aid in determining the types and concentrations of microorganisms during a mission. Still, current microbial monitoring hardware utilizes culture-based methodology which may not detect many medically significant organisms, such as Legionella pneumophila. We hypothesize that evaluation of the ISS environment using non-culture-based technologies would reveal microorganisms not previously reported in spacecraft, allowing for a more complete health assessment. To achieve this goal, a spaceflight experiment, operationally designated as SWAB, was designed to evaluate the DNA from environmental samples collected from ISS and vehicles destined for ISS. Results from initial samples indicate that the sample collection and return procedures were successful. Analysis of these samples using denaturing gradient gel electrophoresis and targeted PCR primers for fungal contaminants is underway. The current results of SWAB and their implication for in-flight molecular analysis of environmental samples will be discussed.

Ott, C. Mark↗
DOE OSTI2022

Exploring the S-process History in the Galactic Disk: Cerium Abundances and Gradients in Open Clusters from the OCCAM/APOGEE Sample

The APOGEE Open Cluster Chemical Abundances and Mapping survey is used to probe the chemical evolution of the s-process element cerium in the Galactic disk. Cerium abundances were derived from measurements of Ce ii lines in the APOGEE spectra using the Brussels Automatic Code for Characterizing High Accuracy Spectra in 218 stars belonging to 42 open clusters. Our results indicate that, in general, for ages < 4 Gyr, younger open clusters have higher [Ce/Fe] and [Ce/α-element] ratios than older clusters. In addition, metallicity segregates open clusters in the [Ce/X]–age plane (where X can be H, Fe, or the α-elements O, Mg, Si, or Ca). These metallicity-dependent relations result in [Ce/Fe] and [Ce/α] ratios with ages that are not universal clocks. Radial gradients of [Ce/H] and [Ce/Fe] ratios in open clusters, binned by age, were derived for the first time, with d[Ce/H]/dR GC being negative, while d[Ce/Fe]/dR GC is positive. [Ce/H] and [Ce/Fe] gradients are approximately constant over time, with the [Ce/Fe] gradient becoming slightly steeper, changing by ~+0.009 dex kpc -1 Gyr -1 . Both the [Ce/H] and [Ce/Fe] gradients are shifted to lower values of [Ce/H] and [Ce/Fe] for older open clusters. The chemical pattern of Ce in open clusters across the Galactic disk is discussed within the context of s-process yields from asymptotic giant branch (AGB) stars, gigayear time delays in Ce enrichment of the interstellar medium, and the strong dependence of Ce nucleosynthesis on the metallicity of its AGB stellar sources.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

An adaptive Hessian approximated stochastic gradient MCMC method

Bayesian approaches have been successfully integrated into training deep neural networks. One popular family is stochastic gradient Markov chain Monte Carlo methods (SG-MCMC), which have gained increasing interest due to their ability to handle large datasets and the potential to avoid overfitting. Although standard SG-MCMC methods have shown great performance in a variety of problems, they may be inefficient when the random variables in the target posterior densities have scale differences or are highly correlated. Here, we present an adaptive Hessian approximated stochastic gradient MCMC method to incorporate local geometric information while sampling from the posterior. The idea is to apply stochastic approximation (SA) to sequentially update a preconditioning matrix at each iteration. The preconditioner possesses second-order information and can guide the random walk of a sampler efficiently. Instead of computing and saving the full Hessian of the log posterior, we use limited memory of the samples and their stochastic gradients to approximate the inverse Hessian-vector multiplication in the updating formula. Moreover, by smoothly optimizing the preconditioning matrix via SA, our proposed algorithm can asymptotically converge to the target distribution with a controllable bias under mild conditions. To reduce the training and testing computational burden, we adopt a magnitude-based weight pruning method to enforce the sparsity of the network. Our method is user-friendly and demonstrates better learning results compared to standard SG-MCMC updating rules. The approximation of inverse Hessian alleviates storage and computational complexities for large dimensional models. Numerical experiments are performed on several problems, including sampling from 2D correlated distribution, synthetic regression problems, and learning the numerical solutions of heterogeneous elliptic PDE. The numerical results demonstrate great improvement in both the convergence rate and accuracy.

97 MATHEMATICS AND COMPUTING↗