Search NASA⌕ Search

SEARCH · Search NASA

Results for “Deep ensemble”

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 271 records · Page 15
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Modeling laser-driven ion acceleration with deep learning

Developments in machine learning promise to ameliorate some of the challenges of modeling complex physical systems through neural-network-based surrogate models. High-intensity, short-pulse lasers can be used to accelerate ions to mega-electronvolt energies, but to model such interactions requires computationally expensive techniques such as particle-in-cell simulations. Multilayer neural networks allow one to take a relatively sparse ensemble of simulations and generate a surrogate model that can be used to rapidly search the parameter space of interest. In this study, we created an ensemble of over 1,000 simulations modeling laser-driven ion acceleration and developed a surrogate to study the resulting parameter space. A neural-network-based approach allows for rapid feature discovery not possible for traditional parameter scans given the computational cost. A notable observation made during this study was the dependence of ion energy on the pre-plasma gradient length scale. While this methodology harbors great promise for ion acceleration, it has ready application to all topics in which large-scale parameter scans are restricted by significant computational cost or relatively large, but sparse, domains.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
NASA NTRS2010

Integrating Enhanced Grace Terrestrial Water Storage Data Into the U.S. and North American Drought Monitors

NASA's Gravity Recovery and Climate Experiment (GRACE) satellites measure time variations nf the Earth's gravity field enabling reliable detection of spatio-temporal variations in total terrestrial water storage (TWS), including ground water. The U.S. and North American Drought Monitors are two of the premier drought monitoring products available to decision-makers for assessing and minimizing drought impacts, but they rely heavily on precipitation indices and do not currently incorporate systematic observations of deep soil moisture and groundwater storage conditions. Thus GRACE has great potential to improve the Drought Monitors hy filling this observational gap. Horizontal, vertical and temporal disaggregation of the coarse-resolution GRACE TWS data has been accomplished by assimilating GRACE TWS anomalies into the Catchment Land Surface Model using ensemble Kalman smoother. The Drought Monitors combine several short-term and long-term drought indices and indicators expressed in percentiles as a reference to their historical frequency of occurrence for the location and time of year in question. To be consistent, we are in the process of generating a climatology of estimated soil moisture and ground water based on m 60-year Catchment model simulation which will subsequently be used to convert seven years of GRACE assimilated fields into soil moisture and groundwater percentiles. for systematic incorporation into the objective blends that constitute Drought Monitor baselines. At this stage we provide a preliminary evaluation of GRACE assimilated Catchment model output against independent datasets including soil moisture observations from Aqua AMSR-E and groundwater level observations from the U.S. Geological Survey's Groundwater Climate Response Network.

Housborg, Rasmus↗
DOE OSTI2022

Application of Dimensionality Reduction in Machine Learning Modeling of CO2 Storage

In this study, we developed deep learning models that are capable of predicting spatio-temporal outputs of CO2 saturation, pressure, and brine production in a 3D saline storage reservoir over 30 years of continuous CO2 injection and a 50-year post-injection timeframe. To improve computational efficiency and maintain performance accuracy, the model framework involves ensembling multi-layer autoencoder networks that provide dimensionality reduction of geologic inputs with fully connected long short-term memory (LSTM) neural networks that generate time-series prediction. This study was presented as poster at the 2022 Carbon Management Project Review Meeting held in Pittsburgh, PA (August 15 – 19, 2022).

Bello, Kolawole↗
DOE OSTI2023

Effect of image resolution on automated classification of chest X-rays

Deep learning (DL) models have received much attention lately for their ability to achieve expert-level performance on the accurate automated analysis of chest X-rays (CXRs). Recently available public CXR datasets include high resolution images, but state-of-the-art models are trained on reduced size images due to limitations on graphics processing unit memory and training time. As computing hardware continues to advance, it has become feasible to train deep convolutional neural networks on high-resolution images without sacrificing detail by downscaling. This study examines the effect of increased resolution on CXR classification performance. We used the publicly available MIMIC-CXR-JPG dataset, comprising 377,110 high resolution CXR images for this study. We applied image downscaling from native resolution to 2048 × 2048 pixels, 1024 × 1024 pixels, 512 × 512 pixels, and 256 × 256 pixels and then we used the DenseNet121 and EfficientNet-B4 DL models to evaluate clinical task performance using these four downscaled image resolutions. We find that while some clinical findings are more reliably labeled using high resolutions, many other findings are actually labeled better using downscaled inputs. We qualitatively verify that tasks requiring a large receptive field are better suited to downscaled low resolution input images, by inspecting effective receptive fields and class activation maps of trained models. Lastly, we show that stacking an ensemble across resolutions outperforms each individual learner at all input resolutions while providing interpretable scale weights, indicating that diverse information is extracted across resolutions.

47 OTHER INSTRUMENTATION↗
DOE OSTI2023

Iterative sampling of expensive simulations for faster deep surrogate training

Deep neural network (DNN) surrogates of expensive physics simulations are enabling a rapid change in the way that common experimental design and analysis tasks are approached. Surrogate models allow simulations to be performed in parallel and separately from downstream tasks, thereby enabling analyses that would be impossible with the simulation in-the-loop; surrogates based on DNNs can effectively emulate diverse non-scalar data of the types collected in fusion and laboratory-astrophysics experiments. The challenge is in training the surrogate model, for which large ensembles of physics simulations must be run, preferably without wasting computational effort on uninteresting simulations. Here, in this paper, we present an iterative sampling scheme that can preferentially propose simulations in interesting regions of parameter space without neglecting unexplored regions, allowing high-quality and wide-ranging surrogate models to be trained using 2–3 times fewer simulations compare to space-filling designs. Our approach uses an explicit importance function defined on the simulation output space, balanced against a measure of simulation density which serves as a proxy for surrogate accuracy. It is easy to implement and can be tuned to find interesting simulations early in the study, allowing surrogates to be trained quickly and refined as new simulations become available; this represents an important step towards the routine generation of deep surrogate models quickly enough to be truly relevant to experimental work.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

Forced Patterns of Sea Level Rise in the Community Earth System Model Large Ensemble From 1920 to 2100

To provide context for observed sea level rise, the forced response (FR) in dynamic sea level (DSL) during the 20th and 21st centuries is examined in the Community Earth System Model Large Ensemble (LE). After accounting for simulation drift, which in the LE is sizable, the DSL FR is found to be complex, both in space and time. Its evolving character is suggested to arise from both the diversity and transient evolution of climate forcing agents and the slow adjustment timescales of the intermediate and deep oceans. Nonetheless, various intervals of spatially coherent change simulated for the recent past and near future are identified, and their characteristics and associated driving mechanisms are identified and discussed. The mid- to late-20th century DSL FR is characterized by an hemispherically asymmetric pattern of change, with depressed rates of rise in the northern oceans. There is also a dipole of change in the Southern Ocean caused by changes in near-surface zonal winds. Through the late 20th and early 21st centuries, a different pattern of rise emerges with elevated rates in the tropics and depressed rates at high latitudes. Zonal and interbasin variations characterize both intervals and involve the pattern and depth of anomalous ocean heat content storage and spatial contrasts in the expansion coefficient tied mainly to base state temperature, with greater rates of rise in warm regions per unit warming. Finally, the relative roles of surface flux and ocean convergence anomalies are examined.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Tree-based algorithms for weakly supervised anomaly detection

Weakly supervised methods have emerged as a powerful tool for model-agnostic anomaly detection at the Large Hadron Collider (LHC). While these methods have shown remarkable performance on specific signatures such as dijet resonances, their application in a more model-agnostic manner requires dealing with a larger number of potentially noisy input features. In this paper, we show that using boosted decision trees as classifiers in weakly supervised anomaly detection gives superior performance compared to deep neural networks. Boosted decision trees are well known for their effectiveness in tabular data analysis. Our results show that they not only offer significantly faster training and evaluation times, but they are also robust to a large number of noisy input features. By using advanced gradient boosted decision trees in combination with ensembling techniques and an extended set of features, we significantly improve the performance of weakly supervised methods for anomaly detection at the LHC. This advance is a crucial step toward a more model-agnostic search for new physics. Published by the American Physical Society 2024

Astronomy & Astrophysics↗
NASA NTRS2013

A Canonical Repsonse of Precipitation Characteristics to Global Warming from CMIP5 Models

In this study, we find from analyses of projections of 14 CMIP5 models a robust, canonical global response in rainfall characteristics to a warming climate. Under a scenario of 1% increase per year of CO2 emission, the model ensemble projects globally more heavy precipitation (+7+/-2.4%/K1), less moderate precipitation (-2.5+/-0.6%/K), more light precipitation (+1.8+/-1.3%/K1), and increased length of dry (no-rain) periods (+4.7+/-2.1%/K). Regionally, a majority of the models project a consistent response with more heavy precipitation over climatologically wet regions of the deep tropics, especially the equatorial Pacific Ocean and the Asian monsoon regions, and more dry periods over the land areas of the subtropics and the tropical marginal convective zones. Our results suggest that increased CO2 emissions induce a global adjustment in circulation and moisture availability manifested in basic changes in global precipitation characteristics, including increasing risks of severe floods and droughts in preferred geographic locations worldwide.

CMIP5↗
NASA NTRSDate not supplied

Model Calibration for Cancer Risk Projections According to Uncertain Data

This paper presents forward and inverse formulations for the calibration of computational models according to uncertain data. Uncertainty in the data might be caused by a poor metrology system, measurement noise, missing or uncontrollable input variables, or by the inability to directly measure the inputs and/or outputs of interest. The forward approach performs the calibration in the space of the model’s output thereby requiring repeated model simulations. Conversely, the inverse approach leverages an ensemble of solutions to an inverse problem in order to perform the calibration in the space of the model’s parameters. As such, the computational demands of the inverse approach are considerably lower. These strategies are applied to the calibration of a radiation model that in-forms cancer risk projections for future deep space missions.

uncertainty quantification↗
NASA NTRS1992

QSO's to B less than 22 and the cosmic x ray background radiation: A fluctuations correlation approach

The contribution of faint Quasi Stellar Objects (QSO's) and other discrete sources to the cosmic X-Ray Background (XRB) is estimated using an optically selected QSO sample in SA 68.2 in conjunction with archived x-ray images from the Einstein Image Proportional Counter (IPC). A new generalization of the 'x-ray image stacking' technique is employed; this new approach tests for a positional 'correlation' between positive x-ray fluctuations in the IPC images, and the locations of the optically selected QSO's. Correlation with thirty SA 68.2 QSO's permits a highly sensitive measurement of the ensemble mean x-ray flux for QSO's with 19 less than B(sub J) less than 22. Although the IPC images are of modest depth, a limiting sensitivity 1.2 times 10 to the minus 14th power erg/sec/sq cm (0.3 to 3.5 keV) is effectively archieved comparable to that of the Rosat medium deep surveys.

Anderson, S. F.↗
NASA NTRS1993

Search for optimal distance spectrum convolutional codes

In order to communicate reliably and to reduce the required transmitter power, NASA uses coded communication systems on most of their deep space satellites and probes (e.g. Pioneer, Voyager, Galileo, and the TDRSS network). These communication systems use binary convolutional codes. Better codes make the system more reliable and require less transmitter power. However, there are no good construction techniques for convolutional codes. Thus, to find good convolutional codes requires an exhaustive search over the ensemble of all possible codes. In this paper, an efficient convolutional code search algorithm was implemented on an IBM RS6000 Model 580. The combination of algorithm efficiency and computational power enabled us to find, for the first time, the optimal rate 1/2, memory 14, convolutional code.

Connor, Matthew C.↗
NASA NTRS2018

Scattering and Extinction by Spherical Particles Immersed in an Absorbing Host Medium

Many applications of electromagnetic scattering involve particles immersed in an absorbing rather than lossless medium, thereby making the conventional scattering theory potentially inapplicable. To analyze this issue quantitatively, we employ the FORTRAN program developed recently on the basis of the first-principles electromagnetic theory to study far-field scattering by spherical particles embedded in an absorbing infinite host medium. We further examine the phenomenon of negative extinction identified recently for monodisperse spheres and uncover additional evidence in favor of its interference origin. We identify the main effects of increasing the width of the size distribution on the ensemble-averaged extinction efficiency factor and show that negative extinction can be eradicated by averaging over a very narrow size distribution. We also analyze, for the first time, the effects of absorption inside the host medium and ensemble averaging on the phase function and other elements of the Stokes scattering matrix. It is shown in particular that increasing absorption significantly suppresses the interference structure and can result in a dramatic expansion of the areas of positive polarization. Furthermore, the phase functions computed for larger effective size parameters can develop a very deep minimum at side-scattering angles bracketed by a strong diffraction peak in the forward direction and a pronounced backscattering maximum.

Far-field electromagnetic scattering; Absorbing ho↗
DOE OSTI2021

Soil Water Potentials (1990-2018) from a calibrated ELM-FATES, and rooting depth analyses scripts, PA-BCI, Panama

This dataset has ELM-FATES simulation outputs and scripts that accompany and reproduce the manuscript, Chitra-Tarak et al. (2021) Hydraulically-vulnerable trees survive on deep-water access during droughts in a tropical forest. New Phytologist. https: //doi.org/10.1111/nph.17464”. This dataset is organized into four parts: (1) source code for the model, Energy Exascale Earth System Land Model-Functionally Assembled Advanced Terrestrial simulator (ELM-FATES) (2) R scripts associated with ELM-FATES simulations: to prepare files for parameter ensemble simulation runs, calibrate the model with observed hydrological fluxes observed at BCI, re-run the model with tuned parameters and obtain long-term soil matric potentials by depth (3) ELM-FATES hydrological simulation outputs at PA-BCI, Panama (1990-2018, daily), and (4) R scripts for inverse rooting depth modeling and post-processing: prepare datasets and inversely model tree rooting depths at BCI, test hypotheses about the relationship of rooting depths with plant hydraulic traits and drought-induced mortality, and reproduce the manuscript itself. n.b. 1. Data needed to run R scripts in (2) and (4) and thus to reproduce the manuscript are sourced from various public datasets. Refer to the data availability section of the accompanying manuscript. 2. This dataset version (prepared on May 12, 2021) matches the revised and published version of the accompanying manuscript (published May 16, 2021) replacing a previous version. In the publication, this dataset is cited as follows: Chitra-Tarak R, Xu C, Aguilar S, Anderson-Teixeira K, Chambers J, Detto M, Faybishenko B, Fisher R, Knox R, Koven C et al. 2020. Soil water potentials (1990–2018) from a calibrated ELM-FATES, and rooting depth analyses scripts, PA-BCI, Panama. 2.0. NGEE Tropics Data Collection. doi: 10.15486/ngt/1696806

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2008

The Impact of Aerosols on Cloud and Precipitation Processes: Cloud-Resolving Model Simulations

Aerosols and especially their effect on clouds are one of the key components of the climate system and the hydrological cycle [Ramanathan et al., 2001]. Yet, the aerosol effect on clouds remains largely unknown and the processes involved not well understood. A recent report published by the National Academy of Science states "The greatest uncertainty about the aerosol climate forcing - indeed, the largest of all the uncertainties about global climate forcing - is probably the indirect effect of aerosols on clouds [NRC, 2001]." The aerosol effect on clouds is often categorized into the traditional "first indirect (i.e., Twomey)" effect on the cloud droplet sizes for a constant liquid water path [Twomey, 1977] and the "semi-direct" effect on cloud coverage [e.g., Ackerman et al ., 2001]." Enhanced aerosol concentrations can also suppress warm rain processes by producing a narrow droplet spectrum that inhibits collision and coalescence processes [e.g., Squires and Twomey, 1961; Warner and Twomey, 1967; Warner, 1968; Rosenfeld, 19991. The aerosol effect on precipitation processes, also known as the second type of aerosol indirect effect [Albrecht, 1989], is even more complex, especially for mixed-phase convective clouds. Table 1 summarizes the key observational studies identifying the microphysical properties, cloud characteristics, thermodynamics and dynamics associated with cloud systems from high-aerosol continental environments. For example, atmospheric aerosol concentrations can influence cloud droplet size distributions, warm-rain process, cold-rain process, cloud-top height, the depth of the mixed phase region, and occurrence of lightning. In addition, high aerosol concentrations in urban environments could affect precipitation variability by providing an enhanced source of cloud condensation nuclei (CCN). Hypotheses have been developed to explain the effect of urban regions on convection and precipitation [van den Heever and Cotton, 2007 and Shepherd, 2005]. Please see Tao et al. (2007) for more detailed description on aerosol impact on precipitation. Recently, a detailed spectral-bin microphysical scheme was implemented into the Goddard Cumulus Ensemble (GCE) model. Atmospheric aerosols are also described using number density size-distribution functions. A spectral-bin microphysical model is very expensive from a computational point of view and has only been implemented into the 2D version of the GCE at the present time. The model is tested by studying the evolution of deep tropical clouds in the west Pacific warm pool region and summertime convection over a mid-latitude continent with different concentrations of CCN: a low "clean" concentration and a high "dirty" concentration. The impact of atmospheric aerosol concentration on cloud and precipitation will be investigated.

Tao, Wei-Kuo↗
NASA NTRS2012

The Impact of Aerosols on Cloud and Precipitation Processes: Cloud-Resolving Model Simulations

Recently, a detailed spectral-bin microphysical scheme was implemented into the Goddard Cumulus Ensemble (GCE) model. Atmospheric aerosols are also described using number density size-distribution functions. A spectral-bin microphysical model is very expensive from a computational point of view and has only been implemented into the 2D version of the GCE at the present time. The model is tested by studying the evolution of deep tropical clouds in the west Pacific warm pool region and summertime convection over a mid-latitude continent with different concentrations of CCN: a low clean concentration and a high dirty concentration. The impact of atmospheric aerosol concentration on cloud and precipitation will be investigated.

cloud-resolving model↗
DOE OSTI2025

Supernova pointing capabilities of DUNE

The determination of the direction of a stellar core collapse via its neutrino emission is crucial for the identification of the progenitor for a multimessenger follow-up. A highly effective method of reconstructing supernova directions within the Deep Underground Neutrino Experiment (DUNE) is introduced. The supernova neutrino pointing resolution is studied by simulating and reconstructing electron-neutrino charged-current absorption on 40 Ar and elastic scattering of neutrinos on electrons. Procedures to reconstruct individual interactions, including a newly developed technique called “brems flipping,” as well as the burst direction from an ensemble of interactions are described. Performance of the burst direction reconstruction is evaluated for supernovae happening at a distance of 10 kpc for a specific supernova burst flux model. The pointing resolution is found to be 3.4 degrees at 68% coverage for a perfect interaction-channel classification and a fiducial mass of 40 kton, and 6.6 degrees for a 10 kton fiducial mass respectively. Assuming a 4% rate of charged-current interactions being misidentified as elastic scattering, DUNE’s burst pointing resolution is found to be 4.3 degrees (8.7 degrees) at 68% coverage.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Supernova pointing capabilities of DUNE

The determination of the direction of a stellar core collapse via its neutrino emission is crucial for the identification of the progenitor for a multimessenger follow-up. A highly effective method of reconstructing supernova directions within the Deep Underground Neutrino Experiment (DUNE) is introduced. The supernova neutrino pointing resolution is studied by simulating and reconstructing electron-neutrino charged-current absorption on Ar 40 and elastic scattering of neutrinos on electrons. Procedures to reconstruct individual interactions, including a newly developed technique called “brems flipping,” as well as the burst direction from an ensemble of interactions are described. Performance of the burst direction reconstruction is evaluated for supernovae happening at a distance of 10 kpc for a specific supernova burst flux model. The pointing resolution is found to be 3.4 degrees at 68% coverage for a perfect interaction-channel classification and a fiducial mass of 40 kton, and 6.6 degrees for a 10 kton fiducial mass respectively. Assuming a 4% rate of charged-current interactions being misidentified as elastic scattering, DUNE’s burst pointing resolution is found to be 4.3 degrees (8.7 degrees) at 68% coverage.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2023

Comparing quantile regression forest and mixture density long short-term memory models for probabilistic post-processing of satellite precipitation-driven streamflow simulations

Abstract. Deep learning (DL) and machine learning (ML) are widely used in hydrological modelling, which plays a critical role in improving the accuracy of hydrological predictions. However, the trade-off between model performance and computational cost has always been a challenge for hydrologists when selecting a suitable model, particularly for probabilistic post-processing with large ensemble members. This study aims to systematically compare the quantile regression forest (QRF) model and countable mixtures of asymmetric Laplacians long short-term memory (CMAL-LSTM) model as hydrological probabilistic post-processors. Specifically, we evaluate their ability in dealing with biased streamflow simulations driven by three satellite precipitation products across 522 nested sub-basins of the Yalong River basin in China. Model performance is comprehensively assessed using a series of scoring metrics from both probabilistic and deterministic perspectives. Our results show that the QRF model and the CMAL-LSTM model are comparable in terms of probabilistic prediction, and their performances are closely related to the flow accumulation area (FAA) of the sub-basin. The QRF model outperforms the CMAL-LSTM model in most sub-basins with smaller FAA, while the CMAL-LSTM model has an undebatable advantage in sub-basins with FAA larger than 60 000 km2 in the Yalong River basin. In terms of deterministic predictions, the CMAL-LSTM model is preferred, especially when the raw streamflow is poorly simulated and used as input. However, setting aside the differences in model performance, the QRF model with 100-member quantiles demonstrates a noteworthy advantage by exhibiting a 50 % reduction in computation time compared to the CMAL-LSTM model with the same ensemble members in all experiments. As a result, this study provides insights into model selection in hydrological post-processing and the trade-offs between model performance and computational efficiency. The findings highlight the importance of considering the specific application scenario, such as the catchment size and the required accuracy level, when selecting a suitable model for hydrological post-processing.

Geology↗