Search NASA⌕ Search

SEARCH · Search NASA

Results for “Distributed training”

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 1,135 records · Page 63

Accelerating Markov Chain Monte Carlo sampling with diffusion models

Global fits of physics models require efficient methods for exploring high-dimensional and/or multimodal posterior functions. We introduce a novel method for accelerating Markov Chain Monte Carlo (MCMC) sampling by pairing a Metropolis-Hastings algorithm with a diffusion model that can draw global samples with the aim of approximating the posterior. We briefly review diffusion models in the context of image synthesis before providing a streamlined diffusion model tailored towards low-dimensional data arrays. We then present our adapted Metropolis-Hastings algorithm which combines local proposals with global proposals taken from a diffusion model that is regularly trained on the samples produced during the MCMC run. Our approach leads to a significant reduction in the number of likelihood evaluations required to obtain an accurate representation of the Bayesian posterior across several analytic functions, as well as for a physical example based on a global fit of parton distribution functions. Our method is extensible to other MCMC techniques, and we briefly compare our method to similar approaches based on normalising flows. A code implementation can be found at https://github.com/NickHunt-Smith/MCMC-diffusion.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Generalized representative structures for atomistic systems

A new method is presented to generate atomic structures that reproduce the essential characteristics of arbitrary material systems, phases, or ensembles. Previous methods allow one to reproduce the essential characteristics (e.g. the chemical disorder) of a large random alloy within a small crystal structure. The ability to generate small representations of random alloys, along with the restriction to crystal systems, results from using the fixed-lattice cluster correlations to describe structural characteristics. A more general description of the structural characteristics of atomic systems is obtained using complete sets of atomic environment descriptors. These are used within for generating representative atomic structures without restriction to fixed lattices. A general data-driven approach is provided here utilizing the atomic cluster expansion (ACE) basis. The N-body ACE descriptors are a complete set of atomic environment descriptors that span both chemical and spatial degrees of freedom and are used within for describing atomic structures. The generalized representative structure (GRS) method presented within generates small atomic structures that reproduce ACE descriptor distributions corresponding to arbitrary structural and chemical complexity. It is shown that systematically improvable representations of crystalline systems on fixed parent lattices, amorphous materials, liquids, and ensembles of atomic structures may be produced efficiently through optimization algorithms. With the GRS method, we highlight reduced representations of atomistic machine-learning training datasets that contain similar amounts of information and small 40–72 atom representations of liquid phases. The ability to use GRS methodology as a driver for informed novel structure generation is also demonstrated. The advantages over other data-driven methods and state-of-the-art methods restricted to high-symmetry systems are highlighted.

atomic cluster expansion↗

DESI luminous red galaxy samples for cross-correlations

We present two galaxy samples, selected from DESI Legacy Imaging Surveys (LS) DR9, with approximately 20,000 square degrees of coverage and spectroscopic redshift distributions designed for cross-correlations such as with CMB lensing, galaxy lensing, and the Sunyaev-Zel'dovich effect. The first sample is identical to the DESI Luminous Red Galaxy (LRG) sample, and the second sample is an extended LRG sample with 2–3 times the DESI LRG density. We present the improved photometric redshifts, tomographic binning and their spectroscopic redshift distributions and imaging systematics weights, and magnification bias coefficients. The catalogs and related data products will be made publicly available. The cosmological constraints using this sample and Planck lensing maps are presented in a companion paper. Here, we also make public the new set of general-purpose photometric redshifts trained using DESI spectroscopic redshifts, which are used in this work, for all galaxies in LS DR9.

79 ASTRONOMY AND ASTROPHYSICS↗

The miniJPAS survey quasar selection – I. Mock catalogues for classification

In this series of papers, we employ several machine learning (ML) methods to classify the point-like sources from the miniJPAS catalogue, and identify quasar candidates. Since no representative sample of spectroscopically confirmed sources exists at present to train these ML algorithms, we rely on mock catalogues. In this first paper, we develop a pipeline to compute synthetic photometry of quasars, galaxies, and stars using spectra of objects targeted as quasars in the Sloan Digital Sky Survey . To match the same depths and signal-to-noise ratio distributions in all bands expected for miniJPAS point sources in the range 17.5 ≤ r < 24, we augment our sample of available spectra by shifting the original r-band magnitude distributions towards the faint end, ensure that the relative incidence rates of the different objects are distributed according to their respective luminosity functions, and perform a thorough modelling of the noise distribution in each filter, by sampling the flux variance either from Gaussian realizations with given widths, or from combinations of Gaussian functions. Finally, we also add in the mocks the patterns of non-detections which are present in all real observations. Although the mock catalogues presented in this work are a first step towards simulated data sets that match the properties of the miniJPAS observations, these mocks can be adapted to serve the purposes of other photometric surveys.

79 ASTRONOMY AND ASTROPHYSICS↗

Deep Convolutional Neural Networks for Distribution System Fault Classification

Faults happen very frequently in distribution systems. Identifying fault types and phases are of critical importance for outage management, fault location, and service restoration. However, this task becomes very challenging due to measurement scarcity in distribution systems. This paper is among the first few that applies deep learning techniques in distribution system fault classification. Specifically, a sequential Convolutional Neural Network(CNN)-based classifier is developed to identify fault buses and phases. The input to the CNN is the steady-state voltage and current data measured at substations. The fault identification is modeled as a multi-label classification problem. Training data under various fault scenarios are obtained in OpenDSS and Gaussian noises are added to mimic measurement errors. A case study in IEEE 13-feeder test system is conducted with single and multiple bus faults scenarios. Numerical results demonstrate the high accuracy and fast computation of the proposed deep CNN-based fault classification.

Tian, Guanyu↗

Design of a Large 5 T Superconducting Magnet for Polarized Target for JLAB Hall B 12 GeV Upgrade Clas12 Detector

Hall B at Jefferson Laboratory (JLAB) will need a 5 T, 78 cm bore polarized target magnet with a field uniformity of {Delta}B/B{sub 0} <; 10{sup -4} in a useful cylinder of the dimensions 0.04 m long x 0.02 m in diameter. The large magnet is designed with a superconducting coil that provides the solenoid with nearly perfect self shielding in order to reduce the fringe field at nearby photo multiplier tubes (PMTs) to less than 3.5 mT. Because the solenoid is also very close to the Clas12 Torus, the nearly perfect shielding provided by the self shielded solenoid greatly reduces force, field, and torque interactions with the six-coil Torus magnet. The solenoid coil consists of 18 coil modules which are made of coils mounted in aluminum plate discs. Each coil module consists of dual double pancake coils with main coils and shield coil partitioned into separate winding cavities in the aluminum plate discs to distribute and reduce radial hoop load and radial coil forces. Each coil module is effectively an enclosed aluminum box and this serves to partition the axial load and thus reduces coil axial forces. Since overall coil forces within each coil module are reduced, this will greatly reduce the number coil training quenches. This is a very important consideration for this solenoid coil because the coil cooling is adiabatic, using in-direct conduction cooling by 4.5 K supercritical helium, which will provide only a relatively small temperature stability margin. Super critical helium is used as per JLAB specification. Detail design of the coil structure, coil assembly, cold mass, and cryogenic control will be presented. The magnet protection system shall be capable of the following features: (1) quench and fault detection, (2) fast discharge of the magnet, (3) limit fault voltages to safe values, (4) monitor interlock signals to prevent unsafe operation, and (5) provide control logic necessary for safe operation of the solenoid. The instrument systems shall be capable of the following features: (1) monitors and display temperatures within the solenoid magnet, (2) measure loads or stress on the magnet suspension, (3) monitor voltages within the solenoid magnet and charging bus, (4) monitor pressures and (5) use data logging system to save all sensor data. These systems will be described and a quench analysis presented.

43 PARTICLE ACCELERATORS↗

Bayesian network ensemble as a multivariate strategy to predict radiation pneumonitis risk

Purpose: Prediction of radiation pneumonitis (RP) has been shown to be challenging due to the involvement of a variety of factors including dose–volume metrics and radiosensitivity biomarkers. Some of these factors are highly correlated and might affect prediction results when combined. Bayesian network (BN) provides a probabilistic framework to represent variable dependencies in a directed acyclic graph. The aim of this study is to integrate the BN framework and a systems’ biology approach to detect possible interactions among RP risk factors and exploit these relationships to enhance both the understanding and prediction of RP. Methods: The authors studied 54 nonsmall-cell lung cancer patients who received curative 3D-conformal radiotherapy. Nineteen RP events were observed (common toxicity criteria for adverse events grade 2 or higher). Serum concentration of the following four candidate biomarkers were measured at baseline and midtreatment: alpha-2-macroglobulin, angiotensin converting enzyme (ACE), transforming growth factor, interleukin-6. Dose-volumetric and clinical parameters were also included as covariates. Feature selection was performed using a Markov blanket approach based on the Koller–Sahami filter. The Markov chain Monte Carlo technique estimated the posterior distribution of BN graphs built from the observed data of the selected variables and causality constraints. RP probability was estimated using a limited number of high posterior graphs (ensemble) and was averaged for the final RP estimate using Bayes’ rule. A resampling method based on bootstrapping was applied to model training and validation in order to control under- and overfit pitfalls. Results: RP prediction power of the BN ensemble approach reached its optimum at a size of 200. The optimized performance of the BN model recorded an area under the receiver operating characteristic curve (AUC) of 0.83, which was significantly higher than multivariate logistic regression (0.77), mean heart dose (0.69), and a pre-to-midtreatment change in ACE (0.66). When RP prediction was made only with pretreatment information, the AUC ranged from 0.76 to 0.81 depending on the ensemble size. Bootstrap validation of graph features in the ensemble quantified confidence of association between variables in the graphs where ten interactions were statistically significant. Conclusions: The presented BN methodology provides the flexibility to model hierarchical interactions between RP covariates, which is applied to probabilistic inference on RP. The authors’ preliminary results demonstrate that such framework combined with an ensemble method can possibly improve prediction of RP under real-life clinical circumstances such as missing data or treatment plan adaptation.

60 APPLIED LIFE SCIENCES↗

Time-Based CAN Intrusion Detection Benchmark

Modern vehicles are complex cyber-physical systems made of hundreds of electronic control units (ECUs) that communicate over controller area networks (CANs). This inherited complexity has expanded the CAN attack surface by the injection of malicious messages that vary their time-based characteristics. To detect these malicious messages, time-based intrusion detection systems (IDS) have been proposed. However, time-based IDS are usually trained and tested on low-fidelity datasets with unrealistic labeled attacks. This makes difficult the task of evaluating, comparing, and validating IDS. Here we detail and benchmark four time-based IDS in a dataset with real and advanced attacks. We found that methods with strong assumptions regarding the distribution of inter-arrival times have lower performance than distribution agnostic based methods. In particular, distribution agnostic based methods outperform distribution based methods at least on $55\%$ in area under the precision-recall (AUC-PR) curve. Our results expand the body of knowledge of CAN time-based IDS by providing details of these methods and reporting their results when tested on datasets with real and advanced attacks. We describe limitations, open challenges, and how lessons learnt from this research can inform the design of deployable time-based IDS in modern vehicles.

Blevins, Deborah↗

A Data-Driven Framework for Automated Detection of Aircraft-Generated Signals in Seismic Array Data Using Machine Learning

Abstract Ground motions associated with aircraft overflights can cover a significant portion of the seismic data collected by shallowly emplaced seismometers, such as new nodal and Distributed Acoustic Sensing systems. This article describes the first published framework for automated detection of aircraft on single channel and multichannel seismic data. The seismic data are converted to spectrograms in a sliding time window and classified as aircraft or nonaircraft in each window using a deep convolutional neural network trained with analyst-labeled data. A majority voting scheme is used to convert the output from the sequence of sliding time windows onto a decision time sequence for each channel and to combine the binary classifications on the decision time sequences across multiple channels. Precision, recall, and F-score are used to quantify the detection performance of the algorithm on nodal data using fourfold time-series cross validation. By applying our framework to data from the Sage Brush Flats nodal array in Southern California, we provide a benchmark performance and demonstrate the advantage of using an array of sensors.

Geochemistry & Geophysics↗

Classification of Broad Absorption Line Quasars with a Convolutional Neural Network

Quasars that exhibit blueshifted, broad absorption lines (BAL QSOs) are an important probe of black hole feedback on galaxy evolution. Yet the presence of BALs is also a complication for large spectroscopic surveys that use quasars as cosmological probes because the BAL features can affect redshift measurements and contaminate information about the matter distribution in the Lyα forest. We present a new BAL QSO catalog for quasars in the Sloan Digital Sky Survey (SDSS) Data Release 14 (DR14). As the SDSS DR14 quasar catalog has over 500,000 quasars, we have developed an automated BAL classifier with a Convolutional Neural Network (CNN). We trained our CNN classifier on the C IV λ 1549 region of a sample of quasars with reliable human classifications, and compared the results to both a dedicated test sample and visual classifications from the earlier SDSS DR12 quasar catalog. Our CNN classifier correctly classifies over 98% of the BAL quasars in the DR12 catalog, which demonstrates comparable reliability to human classification. The disagreements are generally for quasars with lower signal-to-noise ratio spectra and/or weaker BAL features. Our new catalog includes the probability that each quasar is a BAL, the strength, blueshifts and velocity widths of the troughs, and similar information for any Si IV λ 1398 BAL troughs that may be present. We find significant BAL features in 16.8% of all quasars with 1.57 < z < 5.56 in the SDSS DR14 quasar catalog.

79 ASTRONOMY AND ASTROPHYSICS↗

Battery inverter experimental data

The increase in power electronic based generation sources require accurate modeling of inverters. Accurate modeling requires experimental data over wider operation range. We used 30 kW off-the-shelf grid following battery inverter in the experiments. We used controllable AC supply and controllable DC supply to emulate AC and DC side characteristics. The experiments were performed at NREL's Energy Systems Integration Facility. Inverter is tested under 100%, 75%, 50%, 25% load conditions. In the first dataset, for each operating condition, controllable AC source voltage is varied from 0.9 to 1.1 per unit (p.u) with a step value of 0.025 p.u while keeping the frequency at 60 Hz. In the second dataset, under similar load conditions (100%, 75%, 50%, 25% ), the frequency of the controllable AC source voltage was varied from 59 Hz to 61 Hz with a step value of 0.2 Hz. Voltage and frequency range is chosen based on inverter protection. Voltages and currents on DC and AC side are included in the dataset.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Application of Gaussian Process Modeling to Analysis of Functional Unreliability

This paper applies Gaussian Process (GP) modeling to analysis of the functional unreliability of a “passive system.” GPs have been used widely in many ways [1]. The present application uses a GP for emulation of a system simulation code. Such an emulator can be applied in several distinct ways, discussed below. All applications illustrated in this paper have precedents in the literature; the present paper is an application of GP technology to a problem that was originally analyzed [2] using neural networks (NN), and later [3, 4] by a method called “Alternating Conditional Expectations” (ACE). This exercise enables a multifaceted comparison of both the processes and the results. Given knowledge of the range of possible values of key system variables, one could, in principle, quantify functional unreliability by sampling from their joint probability distribution, and performing a system simulation for each sample to determine whether the function succeeded for that particular setting of the variables. Using previously available system simulation codes, such an approach is generally impractical for a plant-scale problem. It has long been recognized, however, that a well-trained code emulator or surrogate could be used in a sampling process to quantify certain performance metrics, even for plant-scale problems. “Response surfaces” were used for this many years ago. But response surfaces are at their best for smoothly varying functions; in regions of parameter space where key system performance metrics may behave in complex ways, or even exhibit discontinuities, response surfaces are not the best available tool. This consideration was one of several that drove the work in [2]. In the present paper, (1) the original quantification of functional unreliability using NN [2], and later ACE [3], is reprised using GP; (2) additional information provided by the GP about uncertainty in the limit surface, generally unavailable in other representations, is discussed; (3) a simple forensic exercise is performed, analogous to the inverse problem of code calibration, but with an accident management spin: given an observation about containment pressure, what can we say about the system variables? References 1. For an introduction to GPs, see (for example) Gaussian Processes for Machine Learning, C. E. Rasmussen and C. K. I. Williams (MIT, 2006). 2. Reliability Quantification of Advanced Reactor Passive Safety Systems, J. J. Vandenkieboom, PhD Thesis (University of Michigan, 1996). 3. Z. Cui, J. C. Lee, J. J. Vandenkieboom, and R. W. Youngblood, “Unreliability Quantification of a Containment Cooling System through ACE and ANN Algorithms,” Trans. Am. Nucl. Soc. 85, 178 (2001). 4. Risk and Safety Analysis of Nuclear Systems, J. C. Lee and N. J. McCormick (Wiley, 2011). See especially §11.2.4.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Uncertainty Quantified Machine Learning for Street Level Flooding Predictions in Norfolk, Virginia

Everyday citizens, emergency responders, and critical infrastructure can be dramatically affected by street flooding caused by the increasing tidal levels and major storm events magnified by climate change. Low-level flooding, such as nuisance flooding, continues to increase in frequency, especially in cities like Norfolk, Virginia, which can expect nearly 200 flooding events by 2050 [1]. Recently, machine learning (ML) models have been leveraged to produce real-time predictions based on local weather and geographic conditions. However, ML models are known to produce unusual results when presented with data that varies from their training set. For decision-makers to determine the trustworthiness of the model's predictions, ML models need to quantify their prediction uncertainty. This study applies Deep Quantile Regression (DQR) to a previously published, LSTM-based model for hourly water depth predictions [2], and analyzes its out-of-distribution (OOD) performance.

Goldenberg, Steven↗

MultiLoad-GAN: A GAN-Based Synthetic Load Group Generation Method Considering Spatial-Temporal Correlations

This paper presents a deep-learning framework, Multi-load Generative Adversarial Network (MultiLoad-GAN), for generating a group of synthetic load profiles (SLPs) simultaneously. The main contribution of MultiLoad-GAN is the capture of spatial-temporal correlations among a group of loads that are served by the same distribution transformer. This enables the generation of a large amount of correlated SLPs required for microgrid and distribution system studies. Here, the novelty and uniqueness of the MultiLoad-GAN framework are three-fold. First, to the best of our knowledge, this is the first method for generating a group of load profiles bearing realistic spatial- temporal correlations simultaneously. Second, two complementary realisticness metrics for evaluating generated load profiles are developed: computing statistics based on domain knowledge and comparing high-level features via a deep-learning classifier. Third, to tackle data scarcity, a novel iterative data augmentation mechanism is developed to generate training samples for enhancing the training of both the classifier and the MultiLoad-GAN model. Simulation results show that MultiLoad- GAN can generate more realistic load profiles than existing approaches, especially in group level characteristics. With little finetuning, MultiLoad-GAN can be readily extended to generate a group of load or PV profiles for a feeder or a service area.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Conditional Pseudo-Reversible Normalizing Flow for Surrogate Modeling in Quantifying Uncertainty Propagation

We introduce a conditional pseudo-reversible normalizing flow (PR-NF) that directly learns conditional probability distributions from noisy physical models to efficiently quantify both forward and inverse uncertainty propagation. Traditional surrogate modeling approaches approximate only the deterministic component of physical models, requiring separate noise characterization and computationally expensive sampling methods for inverse problems. Here, in this work, we develop the conditional PR-NF model to directly learn and efficiently generate samples from the conditional probability density functions (PDFs). The training process utilizes dataset consisting of input-output pairs without requiring prior knowledge about the noise and the function. Once trained, our model efficiently generates samples from conditional PDFs for any input within the training domain. Moreover, the pseudo-reversibility feature allows for the use of fully connected neural network architectures, which simplifies the implementation and enables theoretical analysis. We provide a rigorous convergence analysis of the conditional PR-NF model, showing its ability to converge to the target conditional PDF using the Kullback−Leibler divergence. To demonstrate the effectiveness of our method, we apply it to several benchmark tests and a real-world geologic carbon storage problem.

97 MATHEMATICS AND COMPUTING↗

A numerical method for deriving shape functions of nanoparticles for pair distribution function refinements

We report that in the structural refinement of nanoparticles, discrete atomistic modeling can be used for small nanocrystals (< 15 nm), but becomes computationally unfeasible at larger sizes, where instead unit-cell-based small-box modeling is usually employed. However, the effect of the nanocrystal's shape is often ignored or accounted for with a spherical model regardless of the actual shape due to the complexities of solving and implementing accurate shape effects. Recent advancements have provided a way to determine the shape function directly from a pair distribution function calculated from a discrete atomistic model of any given shape, including both regular polyhedra (e.g. cubes, spheres, octahedra) and anisotropic shapes (e.g. rods, discs, ellipsoids) [Olds et al. (2015). J. Appl. Cryst. 48, 1651–1659], although this approach is still limited to small size regimes due to computational demands. In order to accurately account for the effects of nanoparticle size and shape in small-box refinements, a numerical or analytical description is needed. This article presents a methodology to derive numerical approximations of nanoparticle shape functions by fitting to a training set of known shape functions; the numerical approximations can then be employed on larger sizes yielding a more accurate and physically meaningful refined nanoparticle size. Lastly, the method is demonstrated on a series of simulated and real data sets, and a table of pre-calculated shape function expressions for a selection of common shapes is provided.

36 MATERIALS SCIENCE↗

Controlling extrapolations of nuclear properties with feature selection

Predictions of nuclear properties far from measured data are inherently inaccurate because of uncertainties in our knowledge of nuclear forces and in our treatment of quantum many-body effects in strongly-interacting systems. While the model bias can be directly calculated when experimental data is available, only an estimate can be made in the absence of such measurements. Current approaches to compute the estimated bias quickly lose predictive power when their input variables are taken far from the training region, resulting in uncontrolled uncertainties in applications such as nucleosynthesis simulations. In this letter, we present a novel technique to identify the input variables of machine learning algorithms that can provide robust estimates of model bias. Our process is based on selecting input variables, or features, based on their probability distribution functions across the entire nuclear chart. We illustrate our approach on the problem of quantifying the model bias in nuclear binding energies calculated with Density Functional Theory (DFT). We prove that building model biases with only Z and N as features leads to highly unreliable extrapolations. Conversely, we show that proper feature selection can systematically improve theoretical predictions without increasing uncertainties.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Modeling the aqueous interface of amorphous TiO 2 using deep potential molecular dynamics

Amorphous titanium dioxide (a-TiO 2 ) is widely used as a coating material in applications such as electrochemistry and self-cleaning surfaces where its interface with water has a central role. However, little is known about the structures of the a-TiO 2 surface and aqueous interface, particularly at the microscopic level. In this work, we construct a model of the a-TiO 2 surface via a cut-melt-and-quench procedure based on molecular dynamics simulations with deep neural network potentials (DPs) trained on density functional theory data. After interfacing the a-TiO 2 surface with water, we investigate the structure and dynamics of the resulting system using a combination of DP-based molecular dynamics (DPMD) and ab initio molecular dynamics (AIMD) simulations. Both AIMD and DPMD simulations reveal that the distribution of water on the a-TiO 2 surface lacks distinct layers normally found at the aqueous interface of crystalline TiO 2 , leading to an ~10 times faster diffusion of water at the interface. Bridging hydroxyls (Ti 2 –O b H) resulting from water dissociation decay several times more slowly than terminal hydroxyls (Ti–O w H) due to fast Ti–O w H 2 → Ti–O w H proton exchange events. These results provide a basis for a detailed understanding of the properties of a-TiO 2 in electrochemical environments. Moreover, the procedure of generating the a-TiO 2 -interface employed here is generally applicable to studying the aqueous interfaces of amorphous metal oxides.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗