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 721 records · Page 40

FedOSAA: Improving Federated Learning with One-Step Anderson Acceleration

Federated learning (FL) is a distributed machine learning approach that enables multiple local clients and a central server to collaboratively train a model while keeping the data on their own devices. First-order methods, particularly those incorporating variance reduction techniques, are the most widely used FL algorithms due to their simple implementation and stable performance. However, these methods tend to be slow and require a large number of communication rounds to reach the global minimizer. We propose FedOSAA, a novel approach that preserves the simplicity of first-order methods while achieving the rapid convergence typically associated with second-order methods. Our approach applies one Anderson acceleration (AA) step following classical local updates based on first-order methods with variance reduction, such as FedSVRG and SCAFFOLD, during local training. This AA step is able to leverage curvature information from the history points and gives a new update that approximates the Newton-GMRES direction, thereby significantly improving the convergence. We establish a local linear convergence rate to the global minimizer of FedOSAA for smooth and strongly convex loss functions. Numerical comparisons show that FedOSAA substantially improves the communication and computation efficiency of the original first-order methods, achieving performance comparable to second-order methods like GIANT.

Feng, Xue [University of California, Davis]↗

Simulations and cosmological inference: A statistical model for power spectra means and covariances

We describe an approximate statistical model for the sample variance distribution of the nonlinear matter power spectrum that can be calibrated from limited numbers of simulations. Our model retains the common assumption of a multivariate normal distribution for the power spectrum band powers but takes full account of the (parameter-dependent) power spectrum covariance. The model is calibrated using an extension of the framework in Habib et al. (2007) to train Gaussian processes for the power spectrum mean and covariance given a set of simulation runs over a hypercube in parameter space. We demonstrate the performance of this machinery by estimating the parameters of a power-law model for the power spectrum. Within this framework, our calibrated sample variance distribution is robust to errors in the estimated covariance and shows rapid convergence of the posterior parameter constraints with the number of training simulations.

79 ASTRONOMY AND ASTROPHYSICS↗

Fast and accurate learned multiresolution dynamical downscaling for precipitation

Abstract. This study develops a neural-network-based approach for emulating high-resolution modeled precipitation data with comparable statistical properties but at greatly reduced computational cost. The key idea is to use combination of low- and high-resolution simulations (that differ not only in spatial resolution but also in geospatial patterns) to train a neural network to map from the former to the latter. Specifically, we define two types of CNNs, one that stacks variables directly and one that encodes each variable before stacking, and we train each CNN type both with a conventional loss function, such as mean square error (MSE), and with a conditional generative adversarial network (CGAN), for a total of four CNN variants. We compare the four new CNN-derived high-resolution precipitation results with precipitation generated from original high-resolution simulations, a bilinear interpolater and the state-of-the-art CNN-based super-resolution (SR) technique. Results show that the SR technique produces results similar to those of the bilinear interpolator with smoother spatial and temporal distributions and smaller data variabilities and extremes than the original high-resolution simulations. While the new CNNs trained by MSE generate better results over some regions than the interpolator and SR technique do, their predictions are still biased from the original high-resolution simulations. The CNNs trained by CGAN generate more realistic and physically reasonable results, better capturing not only data variability in time and space but also extremes such as intense and long-lasting storms. The new proposed CNN-based downscaling approach can downscale precipitation from 50 to 12 km in 14 min for 30 years once the network is trained (training takes 4 h using 1 GPU), while the conventional dynamical downscaling would take 1 month using 600 CPU cores to generate simulations at the resolution of 12 km over the contiguous United States.

54 ENVIRONMENTAL SCIENCES↗

Sub-daily Statistical Downscaling of Meteorological Variables Using Neural Networks

A new open source neural network temporal downscaling model is described and tested using CRU-NCEP reanal ysis and CCSM3 climate model output. We downscaled multiple meteorological variables in tandem from monthly to sub-daily time steps while also retaining consistent correlations between variables. We found that our feed forward, error backpropagation approach produced synthetic 6 hourly meteorology with biases no greater than 0.6% across all variables and variance that was accurate within 1% for all variables except atmospheric pressure, wind speed, and precipitation. Correlations between downscaled output and the expected (original) monthly means exceeded 0.99 for all variables, which indicates that this approach would work well for generating atmospheric forcing data consistent with mass and energy conserved GCM output. Our neural network approach performed well for variables that had correlations to other variables of about 0.3 and better and its skill was increased by downscaling multiple correlated variables together. Poor replication of precipitation intensity however required further post-processing in order to obtain the expected probability distribution. The concurrence of precipitation events with expected changes in sub ordinate variables (e.g., less incident shortwave radiation during precipitation events) were nearly as consistent in the downscaled data as in the training data with probabilities that differed by no more than 6%. Our downscaling approach requires training data at the target time step and relies on a weak assumption that climate variability in the extrapolated data is similar to variability in the training data.

54 ENVIRONMENTAL SCIENCES↗

ML-Based Pebble Power Reconstruction for Pebble Bed Reactor Analysis

Pebble power reconstruction has been explored to complement the conventional homogenized modeling approach in pebble bed reactor (PBR) analysis, as detailed heterogeneous geometry calculations are computationally expensive. The random distribution of pebble fuels within the core challenges the application of conventional pin power reconstruction methods. To address this, we introduce a machine learning approach based on the transformer model, composed of encoder and decoder layers, to estimate the flux and power form functions for reconstructing individual pebble neutron fluxes and powers. The homogeneous neutron flux distribution within each spectral zone (SZ) is obtained from finite element solutions of global diffusion or transport calculations. Verification tests demonstrate that the trained transformer model accurately predicts power form functions over a range of conditions, including variations in pebble enrichment, location, type, SZ size, and burnup. In particular, verification using a three-dimensional PBR benchmark with burned pebbles shows good agreement in heterogeneous pebble power distributions between Griffin and Serpent. These results highlight the potential of applying conventional pin power reconstruction approaches to PBR cores with randomly distributed pebbles.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Energy Efficiency Services Sector: Workforce Education and Training Needs

This report provides a baseline assessment of the current state of energy efficiency-related education and training programs and analyzes training and education needs to support expected growth in the energy efficiency services workforce. In the last year, there has been a significant increase in funding for 'green job' training and workforce development (including energy efficiency), through the American Recovery and Reinvestment Act (ARRA). Key segments of the energy efficiency services sector (EESS) have experienced significant growth during the past several years, and this growth is projected to continue and accelerate over the next decade. In a companion study (Goldman et al. 2009), our research team estimated that the EESS will increase two- to four-fold by 2020, to 220,000 person-years of employment (PYE) (low-growth scenario) or up to 380,000 PYE (high-growth scenario), which may represent as many as 1.3 million individuals. In assessing energy efficiency workforce education and training needs, we focus on energy-efficiency services-related jobs that are required to improve the efficiency of residential and nonresidential buildings. Figure ES-1 shows the market value chain for the EESS, sub-sectors included in this study, as well as the types of market players and specific occupations. Our assessment does not include the manufacturing, wholesale, and retail distribution subsectors, or energy efficiency-focused operations and maintenance performed by facility managers.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

QuantumNAT: quantum noise-aware training with noise injection, quantization and normalization

Parameterized Quantum Circuits (PQC) are promising towards quantum advantage on near-term quantum hardware. However, due to the large quantum noises (errors), the performance of PQC models has a severe degradation on real quantum devices. Take Quantum Neural Network (QNN) as an example, the accuracy gap between noise-free simulation and noisy results on IBMQ-Yorktown for MNIST-4 classification is over 60%. Existing noise mitigation methods are general ones without leveraging unique characteristics of PQC; on the other hand, existing PQC work does not consider noise effect. To this end, we present QuantumNAT, a PQC-specific framework to perform noise-aware optimizations in both training and inference stages to improve robustness. We experimentally observe that the effect of quantum noise to PQC measurement outcome is a linear map from noise-free outcome with a scaling and a shift factor. Motivated by that, we propose post-measurement normalization to mitigate the feature distribution differences between noise-free and noisy scenarios. Furthermore, to improve the robustness against noise, we propose noise injection to the training process by inserting quantum error gates to PQC according to realistic noise models of quantum hardware. Lastly, post-measurement quantization is introduced to quantize the measurement outcomes to discrete values, achieving the denoising effect. Extensive experiments on 8 classification tasks using 6 quantum devices demonstrate that QuantumNAT improves accuracy by up to 43%, and achieves over 94% 2-class, 80% 4-class, and 34% 10-class classification accuracy measured on real quantum computers. The code for construction and noise-aware training of PQC is available in the TorchQuantum library.

Wang, Hanrui↗

Development of an Integrated Distribution Management System

This final report details the components, functionality, costs, schedule and benefits of developing an Integrated Distribution Management System (IDMS) for power distribution system operation. The Distribution Automation (DA) and Supervisory Control and Data Acquisition (SCADA) systems used by electric power companies to manage the distribution of electric power to retail energy consumers are vital components of the Nation’s critical infrastructure. Providing electricity is an essential public service and a disruption in that service, if not quickly restored, could threaten the public safety and the Nation’s economic security. Our Nation’s economic prosperity and quality of life have long depended on the essential services that utilities provide; therefore, it is necessary to ensure that electric utilities are able to conduct their operations safely and efficiently. A fully integrated technology of applications is needed to link various remote sensing, communications and control devices with other information tools that help guide Power Distribution Operations personnel. A fully implemented IDMS will provide this, a seamlessly integrated set of applications to raise electric system operating intelligence. IDMS will enhance DA and SCADA through integration of applications such as Geographic Information Systems, Outage Management Systems, Switching Management and Analysis, Operator Training Simulator, and other Advanced Applications, including unbalanced load flow and fault isolation/service restoration. These apps are capable of utilizing and obtaining information from appropriately installed DER, and by integrating disparate systems, the Distribution Operators will benefit from advanced capabilities when analyzing, controlling and operating the electric system.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Extending the SAGA Survey (xSAGA). I. Satellite Radial Profiles as a Function of Host-galaxy Properties

We present "Extending the Satellites Around Galactic Analogs Survey" (xSAGA), a method for identifying low-z galaxies on the basis of optical imaging and results on the spatial distributions of xSAGA satellites around host galaxies. Using spectroscopic redshift catalogs from the SAGA Survey as a training data set, we have optimized a convolutional neural network (CNN) to identify z < 0.03 galaxies from more-distant objects using image cutouts from the DESI Legacy Imaging Surveys. From the sample of >100,000 CNN-selected low-z galaxies, we identify >20,000 probable satellites located between 36–300 projected kpc from NASA-Sloan Atlas central galaxies in the stellar-mass range $9.5\lt \mathrm{log}({M}_{\star }/{M}_{\odot })\lt 11$. We characterize the incompleteness and contamination for CNN-selected samples and apply corrections in order to estimate the true number of satellites as a function of projected radial distance from their hosts. Satellite richness depends strongly on host stellar mass, such that more-massive host galaxies have more satellites, and on host morphology, such that elliptical hosts have more satellites than disky hosts with comparable stellar masses. We also find a strong inverse correlation between satellite richness and the magnitude gap between a host and its brightest satellite. The normalized satellite radial distribution between 36–300 kpc does not depend on host stellar mass, morphology, or magnitude gap. The satellite abundances and radial distributions we measure are in reasonable agreement with predictions from hydrodynamic simulations. Our results deliver unprecedented statistical power for studying satellite galaxy populations and highlight the promise of using machine-learning for extending galaxy samples of wide-area surveys.

79 ASTRONOMY AND ASTROPHYSICS↗

Code Description for "Brief Communication: Monitoring snow depth using small, cheap, and easy-to-deploy ground surface temperature sensors"

Temporally continuous snow depth estimates are vital for understanding changing snow patterns and impacts on permafrost in the Arctic. We train a random forest machine learning model to predict snow depth from variability in ground surface temperature. To our knowledge, this is the first time that small ground surface temperature sensors have been used to estimate snow depth. The model performs well at sites where the model was trained and at pan-arctic evaluation sites (RMSE <= 0.15 m). Small temperature sensors are cheap and easy-to-deploy, so this technique enables spatially distributed and temporally continuous snowpack monitoring to an extent previously infeasible. The model is flexible and can be applied to datasets retroactively to retrieve snow depth estimates at additional sites. This code package includes a *.joblib file of the trained random forest model and a *.ipynb file showing how to clean input data, train the random forest model, and apply the model.

Bachand, Claire↗

A Short-Term and High-Resolution System Load Forecasting Approach Using Support Vector Regression with Hybrid Parameters Optimization

This work proposes an approach for distribution system load forecasting, which aims to provide highly accurate short-term load forecasting with high resolution utilizing a support vector regression (SVR) based forecaster and a two-step hybrid parameters optimization method. Specifically, because the load profiles in distribution systems contain abrupt deviations, a data normalization is designed as the pretreatment for the collected historical load data. Then an SVR model is trained by the load data to forecast the future load. For better performance of SVR, a two-step hybrid optimization algorithm is proposed to determine the best parameters. In the first step of the hybrid optimization algorithm, a designed grid traverse algorithm (GTA) is used to narrow the parameters searching area from a global to local space. In the second step, based on the result of the GTA, particle swarm optimization (PSO) is used to determine the best parameters in the local parameter space. After the best parameters are determined, the SVR model is used to forecast the short-term load deviation in the distribution system.

24 POWER TRANSMISSION AND DISTRIBUTION↗

SU-E-T-23: A Developing Australian Network for Datamining and Modelling Routine Radiotherapy Clinical Data and Radiomics Information for Rapid Learning and Clinical Decision Support

Purpose: Large amounts of routine radiotherapy (RT) data are available, which can potentially add clinical evidence to support better decisions. A developing collaborative Australian network, with a leading European partner, aims to validate, implement and extend European predictive models (PMs) for Australian practice and assess their impact on future patient decisions. Wider objectives include: developing multi-institutional rapid learning, using distributed learning approaches; and assessing and incorporating radiomics information into PMs. Methods: Two initial standalone pilots were conducted; one on NSCLC, the other on larynx, patient datasets in two different centres. Open-source rapid learning systems were installed, for data extraction and mining to collect relevant clinical parameters from the centres’ databases. The European DSSs were learned (“training cohort”) and validated against local data sets (“clinical cohort”). Further NSCLC studies are underway in three more centres to pilot a wider distributed learning network. Initial radiomics work is underway. Results: For the NSCLC pilot, 159/419 patient datasets were identified meeting the PM criteria, and hence eligible for inclusion in the curative clinical cohort (for the larynx pilot, 109/125). Some missing data were imputed using Bayesian methods. For both, the European PMs successfully predicted prognosis groups, but with some differences in practice reflected. For example, the PM-predicted good prognosis NSCLC group was differentiated from a combined medium/poor prognosis group (2YOS 69% vs. 27%, p<0.001). Stage was less discriminatory in identifying prognostic groups. In the good prognosis group two-year overall survival was 65% in curatively and 18% in palliatively treated patients. Conclusion: The technical infrastructure and basic European PMs support prognosis prediction for these Australian patient groups, showing promise for supporting future personalized treatment decisions, improved treatment quality and potential practice changes. The early indications from the distributed learning and radiomics pilots strengthen this. Improved routine patient data quality should strengthen such rapid learning systems.

60 APPLIED LIFE SCIENCES↗

A Hybrid Semi-supervised Classification Scheme for Mining Multisource Geospatial Data

Supervised learning methods such as Maximum Likelihood (ML) are often used in land cover (thematic) classification of remote sensing imagery. ML classifier relies exclusively on spectral characteristics of thematic classes whose statistical distributions (class conditional probability densities) are often overlapping. The spectral response distributions of thematic classes are dependent on many factors including elevation, soil types, and ecological zones. A second problem with statistical classifiers is the requirement of large number of accurate training samples (10 to 30 |dimensions|), which are often costly and time consuming to acquire over large geographic regions. With the increasing availability of geospatial databases, it is possible to exploit the knowledge derived from these ancillary datasets to improve classification accuracies even when the class distributions are highly overlapping. Likewise newer semi-supervised techniques can be adopted to improve the parameter estimates of statistical model by utilizing a large number of easily available unlabeled training samples. Unfortunately there is no convenient multivariate statistical model that can be employed for mulitsource geospatial databases. In this paper we present a hybrid semi-supervised learning algorithm that effectively exploits freely available unlabeled training samples from multispectral remote sensing images and also incorporates ancillary geospatial databases. We have conducted several experiments on real datasets, and our new hybrid approach shows over 25 to 35% improvement in overall classification accuracy over conventional classification schemes.

54 ENVIRONMENTAL SCIENCES↗

SU-D-201-07: A Survey of Radiation Oncology Residents’ Training and Preparedness to Lead Patient Safety Programs in Clinics

Purpose: Safety and quality has garnered increased attention in radiation oncology, and physicians and physicists are ideal leaders of clinical patient safety programs. However, it is not clear whether residency programs incorporate formal patient safety training and adequately equip residents to assume this leadership role. A national survey was conducted to evaluate medical and physics residents’ exposure to safety topics and their confidence with the skills required to lead clinical safety programs. Methods: Radiation oncology residents were identified in collaboration with ARRO and AAPM. The survey was released in February 2016 via email using REDCap. This included questions about exposure to safety topics, confidence leading safety programs, and interest in training opportunities (i.e. workshops). Residents rated their exposure, skills, and confidence on 4 or 5-point scales. Medical and physics residents responses were compared using chi-square tests. Results: Responses were collected from 56 of 248 (22%) physics and 139 of 690 (20%) medical residents. More than two thirds of all residents had no or only informal exposure to incident learning systems (ILS), root cause analysis (RCA), failure mode and effects analysis (FMEA), and the concept of human factors engineering (HFE). Likewise, 63% of residents had not heard of RO-ILS. Response distributions were similar, however more physics residents had formal exposure to FMEA (p<0.0001) and felt they were adequately trained to lead FMEAs in clinic (p<0.001) than medical residents. Only 36% of residents felt their patient safety training was adequate, and 58% felt more training would benefit their education. Conclusion: These results demonstrate that, despite increasing desire for patient safety training, medical and physics residents’ exposure to relevant concepts is low. Physics residents had more exposure to FMEA than medical residents, and were more confident in leading FMEA. This suggests that increasing resident exposure to specific topics may increase their confidence.

60 APPLIED LIFE SCIENCES↗

Situational Awareness in Distribution Grid Using Micro-PMU Data: A Machine Learning Approach

The recent development of distribution-level phasor measurement units, a.k.a. micro-PMUs, has been an important step towards achieving situational awareness in power distribution networks. The challenge however is to transform the large amount of data that is generated by micro-PMUs to actionable information and then match the information to use cases with practical value to system operators. This open problem is addressed in this paper. First, we introduce a novel data-driven event detection technique to pick out valuable portion of data from extremely large raw micro-PMU data. Subsequently, a datadriven event classifier is developed to effectively classify power quality events. Importantly, we use field expert knowledge and utility records to conduct an extensive data-driven event labeling. Moreover, certain aspects from event detection analysis are adopted as additional features to be fed into the classifier model. In this regard, a multi-class support vector machine (multi-SVM) classifier is trained and tested over 15 days of real-world data from two micro-PMUs on a distribution feeder in Riverside, CA. In total, we analyze 1.2 billion measurement points, and 10,700 events. Here, the effectiveness of the developed event classifier is compared with prevalent multi-class classification methods, including k-nearest neighbor method as well as decision-tree method. Importantly, two real-world use-cases are presented for the proposed data analytics tools, including remote asset monitoring and distribution-level oscillation analysis.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Shared virtual environments for aerospace training

Virtual environments have the potential to significantly enhance the training of NASA astronauts and ground-based personnel for a variety of activities. A critical requirement is the need to share virtual environments, in real or near real time, between remote sites. It has been hypothesized that the training of international astronaut crews could be done more cheaply and effectively by utilizing such shared virtual environments in the early stages of mission preparation. The Software Technology Branch at NASA's Johnson Space Center has developed the capability for multiple users to simultaneously share the same virtual environment. Each user generates the graphics needed to create the virtual environment. All changes of object position and state are communicated to all users so that each virtual environment maintains its 'currency.' Examples of these shared environments will be discussed and plans for the utilization of the Department of Defense's Distributed Interactive Simulation (DIS) protocols for shared virtual environments will be presented. Finally, the impact of this technology on training and education in general will be explored.

Loftin, R. Bowen↗

EV Forecasting-Based Model Predictive Control for Distribution System Congestion Mitigation

The uncoordinated charging of electric vehicles (EVs) in time and space brings congestion issues to the distribution network. This paper proposes an EV charging demand forecasting-based model predictive control (MPC) method for distribution system congestion management. To effectively forecast the time-series EV station charging demand, a hybrid forecasting model that integrates the long short-term memory network (LSTM) and Transformer is proposed. The Transformer-LSTM model is trained using a one-year real historical charging dataset of EV stations to forecast future charging demand in 15-minute intervals. This informs the MPC for distribution network congestion management and minimization of PV curtailment. Numerical results carried out on the modified IEEE 123-bus distribution system demonstrate that the proposed method can effectively resolve line congestion issues through EV smart charging and PV curtailment while outperforming other benchmarks.

ADVANCED PROPULSION SYSTEMS,SOLAR ENERGY↗

Correlation between model observer and human observer performance in CT imaging when lesion location is uncertain

Purpose: The purpose of this study was to investigate the correlation between model observer and human observer performance in CT imaging for the task of lesion detection and localization when the lesion location is uncertain.Methods: Two cylindrical rods (3-mm and 5-mm diameters) were placed in a 35 × 26 cm torso-shaped water phantom to simulate lesions with −15 HU contrast at 120 kV. The phantom was scanned 100 times on a 128-slice CT scanner at each of four dose levels (CTDIvol = 5.7, 11.4, 17.1, and 22.8 mGy). Regions of interest (ROIs) around each lesion were extracted to generate images with signal-present, with each ROI containing 128 × 128 pixels. Corresponding ROIs of signal-absent images were generated from images without lesion mimicking rods. The location of the lesion (rod) in each ROI was randomly distributed by moving the ROIs around each lesion. Human observer studies were performed by having three trained observers identify the presence or absence of lesions, indicating the lesion location in each image and scoring confidence for the detection task on a 6-point scale. The same image data were analyzed using a channelized Hotelling model observer (CHO) with Gabor channels. Internal noise was added to the decision variables for the model observer study. Area under the curve (AUC) of ROC and localization ROC (LROC) curves were calculated using a nonparametric approach. The Spearman's rank order correlation between the average performance of the human observers and the model observer performance was calculated for the AUC of both ROC and LROC curves for both the 3- and 5-mm diameter lesions.Results: In both ROC and LROC analyses, AUC values for the model observer agreed well with the average values across the three human observers. The Spearman's rank order correlation values for both ROC and LROC analyses for both the 3- and 5-mm diameter lesions were all 1.0, indicating perfect rank ordering agreement of the figures of merit (AUC) between the average performance of the human observers and the model observer performance.Conclusions: In CT imaging of different sizes of low-contrast lesions (−15 HU), the performance of CHO with Gabor channels was highly correlated with human observer performance for the detection and localization tasks with uncertain lesion location in CT imaging at four clinically relevant dose levels. This suggests the ability of Gabor CHO model observers to meaningfully assess CT image quality for the purpose of optimizing scan protocols and radiation dose levels in detection and localization tasks for low-contrast lesions.

61 RADIATION PROTECTION AND DOSIMETRY↗