Search NASA⌕ Search

SEARCH · Search NASA

Results for “Randomized methods”

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 613 records · Page 34
All sourcesNASA NTRSDOE OSTI
DOE OSTI2015

A stochastic model and Monte Carlo algorithm for fluctuation-induced H{sub 2} formation on the surface of interstellar dust grains

A stochastic algorithm for simulation of fluctuation-induced kinetics of H{sub 2} formation on grain surfaces is suggested as a generalization of the technique developed in our recent studies [1] where this method was developed to describe the annihilation of spatially separate electrons and holes in a disordered semiconductor. The stochastic model is based on the spatially inhomogeneous, nonlinear integro-differential Smoluchowski equations with random source term. In this paper we derive the general system of Smoluchowski type equations for the formation of H{sub 2} from two hydrogen atoms on the surface of interstellar dust grains with physisorption and chemisorption sites. We focus in this study on the spatial distribution, and numerically investigate the segregation in the case of a source with a continuous generation in time and randomly distributed in space. The stochastic particle method presented is based on a probabilistic interpretation of the underlying process as a stochastic Markov process of interacting particle system in discrete but randomly progressed time instances. The segregation is analyzed through the correlation analysis of the vector random field of concentrations which appears to be isotropic in space and stationary in time.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2014

Photometric redshift analysis in the Dark Energy Survey Science Verification data

In this study, we present results from a study of the photometric redshift performance of the Dark Energy Survey (DES), using the early data from a Science Verification period of observations in late 2012 and early 2013 that provided science-quality images for almost 200 sq. deg. at the nominal depth of the survey. We assess the photometric redshift (photo-z) performance using about 15 000 galaxies with spectroscopic redshifts available from other surveys. These galaxies are used, in different configurations, as a calibration sample, and photo-z's are obtained and studied using most of the existing photo-z codes. A weighting method in a multidimensional colour–magnitude space is applied to the spectroscopic sample in order to evaluate the photo-z performance with sets that mimic the full DES photometric sample, which is on average significantly deeper than the calibration sample due to the limited depth of spectroscopic surveys. In addition, empirical photo-z methods using, for instance, artificial neural networks or random forests, yield the best performance in the tests, achieving core photo-z resolutions σ68 ~ 0.08. Moreover, the results from most of the codes, including template-fitting methods, comfortably meet the DES requirements on photo-z performance, therefore, providing an excellent precedent for future DES data sets.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2023

Efficient QAOA Optimization using Directed Restarts and Graph Lookup

Variational Quantum Algorithms (VQA) aim to enhance the capabilities of Noisy Intermediate-Scale Quantum (NISQ) devices. These algorithms utilize parameterized circuits and classical optimizers to iteratively execute circuits with varying parameters. However, VQA faces computational overheads due to repeated iterations and random restarts. Prior work suggests using basic sub-graphs to transfer parameters for the input graph, reducing optimizer overheads but limiting applicability to structured regular graphs. In real-world applications, random irregular graphs are common, and existing methods are not scalable or practical for such graphs. This paper presents a framework that aims to improve random irregular graphs in VQA. The framework uses graph similarity and important features like total edge counts, average edge counts, and variance. It follows an iterative process to choose basis sub-graphs from a small database and adjust parameters accordingly. Classical optimizers then utilize these parameters to determine when to restart and perform gradient descent. This approach increases the chances of reaching global maximum points.

Wang, Meng↗
DOE OSTI2024

Improved Subseasonal Forecasting of Extreme Polar Vortices Using Machine Learning

Our research was focused on forecasting the position and shape of the winter stratospheric polar vortex at a subseasonal timescale of 15 days in advance. To achieve this, we employed both statistical and neural network machine learning techniques. The analysis was performed on 42 winter seasons of reanalysis data provided by NASA giving us a total of 6,342 days of data. The state of the polar vortex for determined by using geometric moments to calculate the centroid latitude and the aspect ratio of an ellipse fit onto the vortex. Timeseries for thirty additional precursors were calculated to help improve the predictive capabilities of the algorithm. Feature importance of these precursors was performed using random forest to measure the predictive importance and the ideal number of precursors. Then, using the precursors identified as important, various statistical methods were tested for predictive accuracy with random forest and nearest neighbor performing the best. An echo state network, a type of recurrent neural network that features sparsely connected hidden layer and a reduced number of trainable parameters that allows for rapid training and testing, was also implemented for the forecasting problem. Hyperparameter tuning was performed for each methods using a subset of the training data. The algorithms were trained and tuned on the first 41 years of data, then tested for accuracy on the final year. In general, the centroid latitude of the polar vortex proved easier to predict than the aspect ratio across all algorithms. Random forest outperformed other statistical forecasting algorithms overall but struggled to predict extreme values. Forecasting from echo state network suggested a strong predictive capability past 15 days, but further work is required to fully realize the potential of recurrent neural network approaches.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2012

ACCELERATION OF VERY SMALL DUST GRAINS DUE TO RANDOM CHARGE FLUCTUATIONS

We study the acceleration of very small dust grains including polycyclic aromatic hydrocarbons arising from electrostatic interactions of dust grains that have charge fluctuating randomly in time. Random charge fluctuations of very small grains due to discrete charging events (i.e., sticking collisions with electrons and ions in plasma, and emission of photoelectrons by UV photons) are simulated using the Monte Carlo (MC) method. The motion of dust grains in randomly fluctuating electric fields induced by surrounding charged grains is studied using MC simulations. We identify the acceleration induced by random charge fluctuations as a dominant acceleration mechanism for very small grains in the diffuse interstellar medium (ISM). We find that this acceleration mechanism is efficient for environments with a low degree of ionization (i.e., large Debye length), where charge fluctuations are slow but have a large amplitude. The implications of the present acceleration mechanism for grain coagulation and shattering in the diffuse ISM and dark clouds are also discussed.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2011

A new class of accelerated kinetic Monte Carlo algorithms

Kinetic ( aka dynamic) Monte Carlo (KMC) is a powerful method for numerical simulations of time dependent evolution applied in a wide range of contexts including biology, chemistry, physics, nuclear sciences, financial engineering, etc. Generally, in a KMC the time evolution takes place one event at a time, where the sequence of events and the time intervals between them are selected (or sampled) using random numbers. While details of the method implementation vary depending on the model and context, there exist certain common issues that limit KMC applicability in almost all applications. Among such is the notorious 'flicker problem' where the same states of the systems are repeatedly visited but otherwise no essential evolution is observed. In its simplest form the flicker problem arises when two states are connected to each other by transitions whose rates far exceed the rates of all other transitions out of the same two states. In such cases, the model will endlessly hop between the two states otherwise producing no meaningful evolution. In most situation of practical interest, the trapping cluster includes more than two states making the flicker somewhat more difficult to detect and to deal with. Several methods have been proposed to overcome or mitigate the flicker problem, exactly or approximately. Of the exact methods, the one proposed by Novotny is perhaps most relevant to our research. Novotny formulates the problem of escaping from a trapping cluster as a Markov system with absorbing states. Given an initial state inside the cluster, it is in principle possible to solve the Master Equation for the time dependent probabilities to find the walker in a given state (transient or absorbing) of the cluster at any time in the future. Novotny then proceeds to demonstrate implementation of his general method to trapping clusters containing the initial state plus one or two transient states and all of their absorbing states. Similar methods have been subsequently proposed in [refs] but applied in a different context. The most serious deficiency of the earlier methods is that size of the trapping cluster size is fixed and often too small to bring substantial simulation speedup. Furthermore, the overhead associated with solving for the probability distribution on the trapping cluster sometimes makes such simulations less efficient than the standard KMC. Here we report on a general and exact accelerated kinetic Monte Carlo algorithm generally applicable to arbitrary Markov models1. Two different implementations are attempted both based on incremental expansion of trapping sub-set of Markov states: (1) numerical solution of the Master Equation with absorbing states and (2) incremental graph reduction followed by randomization. Of the two implementations, the 2nd one performs better allowing, for the first time, to overcome trapping basins spanning several million Markov states. The new method is used for simulations of anomalous diffusion on a 2D substrate and of the kinetics of diffusive 1st order phase transformations in binary alloys. Depending on temperature and (alloy) super-saturation conditions, speedups of 3 to 7 orders of magnitude are demonstrated, with no compromise of simulation accuracy.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2019

Effects of Alcohol Administered with Flibanserin in Healthy Premenopausal Women: A Randomized, Double-Blind, Single-Dose Crossover Study

Abstract Introduction Flibanserin is approved in the United States and Canada for the treatment of hypoactive sexual desire disorder in premenopausal women. Aim The purpose of this trial was to evaluate the safety of concomitant administration of flibanserin with alcohol. Methods In this single-center, randomized, double-blind, single-dose, crossover study, participants were randomly assigned to 1 of 12 sequences to receive each of 7 treatments: flibanserin 100 mg or placebo with ethanol 0.2 g/kg, 0.4 g/kg, or 0.6 g/kg, or flibanserin 100 mg only. Treatments were administered using a worst-case approach that included morning dosing and consumption of alcohol within 10 minutes. Main Outcome Measure The primary end point was the proportion of participants who experienced dizziness, syncope, or hypotension. Safety end points included orthostatic vital signs. Results The study included 96 premenopausal women (mean age 31 ± 8 years). The incidence of dizziness for ethanol + flibanserin was 39.8% for ethanol 0.6 g/kg, 34.1% for 0.4 g/kg, and 27.4% for 0.2 g/kg compared with 31.1% for flibanserin without ethanol. Based on the available vital signs data, there was no effect of ethanol concentration on orthostatic blood pressure, vertigo, or hypotension; no instances of syncope were observed. The overall incidence of adverse events (AEs) was similar when flibanserin was administered alone (96.7%) or with ethanol (90.5–97.6%). Clinical Implications Consumption of the tested amounts of alcohol (0.2–0.6 g/kg) does not have an additive effect on the AE profile of flibanserin 100 mg in healthy premenopausal women. Strengths & Limitations Strengths include the study population (premenopausal women, as indicated for flibanserin) and range of ethanol doses. Limitations include the morning dosing of study medication, which is inconsistent with the bedtime dosing recommended for flibanserin, and the method of handling missing vital sign measurements. Conclusion Co-administration of flibanserin 100 mg with varying doses of ethanol resulted in few AEs of special interest, with no notable alcohol dose response. However, a significantly greater percentage of participants administered flibanserin with 0.6 g/kg and 0.4 g/kg of alcohol were characterized as “Participants in Whom Standing Blood Pressure Was Not Obtained” compared with participants administered flibanserin alone.

Simon, James A.↗
DOE OSTI2012

Automatic tracking of implanted fiducial markers in cone beam CT projection images

Purpose: This paper describes a novel method for simultaneous intrafraction tracking of multiple fiducial markers. Although the proposed method is generic and can be adopted for a number of applications including fluoroscopy based patient position monitoring and gated radiotherapy, the tracking results presented in this paper are specific to tracking fiducial markers in a sequence of cone beam CT projection images. Methods: The proposed method is accurate and robust thanks to utilizing the mean shift and random sampling principles, respectively. The performance of the proposed method was evaluated with qualitative and quantitative methods, using data from two pancreatic and one prostate cancer patients and a moving phantom. The ground truth, for quantitative evaluation, was calculated based on manual tracking preformed by three observers. Results: The average dispersion of marker position error calculated from the tracking results for pancreas data (six markers tracked over 640 frames, 3840 marker identifications) was 0.25 mm (at iscoenter), compared with an average dispersion for the manual ground truth estimated at 0.22 mm. For prostate data (three markers tracked over 366 frames, 1098 marker identifications), the average error was 0.34 mm. The estimated tracking error in the pancreas data was < 1 mm (2 pixels) in 97.6% of cases where nearby image clutter was detected and in 100.0% of cases with no nearby image clutter. Conclusions: The proposed method has accuracy comparable to that of manual tracking and, in combination with the proposed batch postprocessing, superior robustness. Marker tracking in cone beam CT (CBCT) projections is useful for a variety of purposes, such as providing data for assessment of intrafraction motion, target tracking during rotational treatment delivery, motion correction of CBCT, and phase sorting for 4D CBCT.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2022

Effective Field Theory of Random Quantum Circuits

Quantum circuits have been widely used as a platform to simulate generic quantum many-body systems. In particular, random quantum circuits provide a means to probe universal features of many-body quantum chaos and ergodicity. Some such features have already been experimentally demonstrated in noisy intermediate-scale quantum (NISQ) devices. On the theory side, properties of random quantum circuits have been studied on a case-by-case basis and for certain specific systems, and a hallmark of quantum chaos—universal Wigner–Dyson level statistics—has been derived. This work develops an effective field theory for a large class of random quantum circuits. The theory has the form of a replica sigma model and is similar to the low-energy approach to diffusion in disordered systems. The method is used to explicitly derive the universal random matrix behavior of a large family of random circuits. In particular, we rederive the Wigner–Dyson spectral statistics of the brickwork circuit model by Chan, De Luca, and Chalker [Phys. Rev. X 8, 041019 (2018)] and show within the same calculation that its various permutations and higher-dimensional generalizations preserve the universal level statistics. Finally, we use the replica sigma model framework to rederive the Weingarten calculus, which is a method of evaluating integrals of polynomials of matrix elements with respect to the Haar measure over compact groups and has many applications in the study of quantum circuits. The effective field theory derived here provides both a method to quantitatively characterize the quantum dynamics of random Floquet systems (e.g., calculating operator and entanglement spreading) and a path to understanding the general fundamental mechanism behind quantum chaos and thermalization in these systems.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2023

Posiform planting: generating QUBO instances for benchmarking

We are interested in benchmarking both quantum annealing and classical algorithms for minimizing quadratic unconstrained binary optimization (QUBO) problems. Such problems are NP-hard in general, implying that the exact minima of randomly generated instances are hard to find and thus typically unknown. While brute forcing smaller instances is possible, such instances are typically not interesting due to being too easy for both quantum and classical algorithms. In this contribution, we propose a novel method, called posiform planting , for generating random QUBO instances of arbitrary size with known optimal solutions, and use those instances to benchmark the sampling quality of four D-Wave quantum annealers utilizing different interconnection structures (Chimera, Pegasus, and Zephyr hardware graphs) and the simulated annealing algorithm. Posiform planting differs from many existing methods in two key ways. It ensures the uniqueness of the planted optimal solution, thus avoiding groundstate degeneracy, and it enables the generation of QUBOs that are tailored to a given hardware connectivity structure, provided that the connectivity is not too sparse. Posiform planted QUBOs are a type of 2-SAT boolean satisfiability combinatorial optimization problems. Our experiments demonstrate the capability of the D-Wave quantum annealers to sample the optimal planted solution of combinatorial optimization problems with up to 5, 627 qubits.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

A spatially regularized detector for emergent/re-emergent disease outbreaks

Early detection of outbreaks caused by emergent pathogens, using epidemiological surveillance data i.e., daily case counts, is difficult. This is because the data tend to be noisy during the early epoch of the outbreak. In contrast, the spread-rate of the disease tends to be well-behaved, as it depends only on the mixing patterns of the population and the characteristics of the pathogen, neither of which behave erratically in space-time. In this report, we explore whether the spread-rate can be used for epidemiological surveillance, conditional on case count data. Estimating the spread-rate from case count data allows us to exploit exogenous information, e.g., incubation period distributions etc., which can considerably smooth out any erratic temporal behavior. Further, epidemiological dynamics are spatially correlated, and if case counts are available for multiple areal units e.g., counties, these correlations could potentially be used to suppress noise in the early epoch data. These exogenous information and structure are not exploited by conventional syndromic surveillance detectors to extract a well-behaved latent variable for monitoring purposes. The technical challenge lies in the estimation of the spread-rate field jointly over a collection of areal units; further, the spread-rate varies over time. We develop a method based on mean-field variational inference to approximately estimate the spread-rate field, using a Gaussian Random Field Model for spatial regularization. The method is tested on the estimation of spread-rate in the thirty-three counties of New Mexico and detect the arrival of the Fall 2020 COVID-19 wave in September 2020. We find that the method is scalable, but underestimates the uncertainty in the estimated spread-rate field. We detect the arrival of the Fall 2020 wave a week ahead of conventional syndromic surveillance algorithms, but our simplistic detection algorithm, based on simple anomaly detection, suffers from a high false positive rate, similar to conventional detectors.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2015

Explicitly correlated ring-coupled-cluster-doubles theory

The connection between the random-phase approximation and the ring-coupled-cluster-doubles method bridges the gap between density-functional and wave-function theories and the importance of the random-phase approximation lies in both its broad applicability and this linking role in electronic-structure theory. In this contribution, we present an explicitly correlated approach to the random-phase approximation, based on the direct ring-coupled-cluster-doubles ansatz, which overcomes the problem of slow basis-set convergence, inherent to the random-phase approximation. Benchmark results for a test set of 106 molecules and a selection of 10 organic complexes from the S22 test set demonstrate that convergence to within 99% of the basis-set limit is reached for triple-zeta basis sets for atomisation energies, while quadruple-zeta basis sets are required for interaction energies. Corrections due to single excitations into the complementary auxiliary space reduce the basis-set incompleteness error by one order of magnitude, while contributions due to the coupling of conventional and geminal amplitudes are in general negligible. We find that a non-iterative explicitly correlated correction to first order in perturbation theory exhibits the best ratio of accuracy to computational cost.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2014

Data-Driven Benchmarking of Building Energy Efficiency Utilizing Statistical Frontier Models

Frontier methods quantify the energy efficiency of buildings by forming an efficient frontier (best-practice technology) and by comparing all buildings against that frontier. Because energy consumption fluctuates over time, the efficiency scores are stochastic random variables. Existing applications of frontier methods in energy efficiency either treat efficiency scores as deterministic values or estimate their uncertainty by resampling from one set of measurements. Availability of smart meter data (repeated measurements of energy consumption of buildings) enables using actual data to estimate the uncertainty in efficiency scores. Additionally, existing applications assume a linear form for an efficient frontier; i.e.,they assume that the best-practice technology scales up and down proportionally with building characteristics. However, previous research shows that buildings are nonlinear systems. This paper proposes a statistical method called stochastic energy efficiency frontier (SEEF) to estimate a bias-corrected efficiency score and its confidence intervals from measured data. The paper proposes an algorithm to specify the functional form of the frontier, identify the probability distribution of the efficiency score of each building using measured data, and rank buildings based on their energy efficiency. To illustrate the power of SEEF, this paper presents the results from applying SEEF on a smart meter data set of 307 residential buildings in the United States. SEEF efficiency scores are used to rank individual buildings based on energy efficiency, to compare subpopulations of buildings, and to identify irregular behavior of buildings across different time-of-use periods. SEEF is an improvement to the energy-intensity method (comparing kWh/sq.ft.): whereas SEEF identifies efficient buildings across the entire spectrum of building sizes, the energy-intensity method showed bias toward smaller buildings. The results of this research are expected to assist researchers and practitioners compare and rank (i.e.,benchmark) buildings more robustly and over a wider range of building types and sizes. Eventually, doing so is expected to result in improved resource allocation in energy-efficiency programs.

Kavousian, A↗
DOE OSTI2015

Local polynomial chaos expansion for linear differential equations with high dimensional random inputs

In this paper we present a localized polynomial chaos expansion for partial differential equations (PDE) with random inputs. In particular, we focus on time independent linear stochastic problems with high dimensional random inputs, where the traditional polynomial chaos methods, and most of the existing methods, incur prohibitively high simulation cost. Furthermore, the local polynomial chaos method employs a domain decomposition technique to approximate the stochastic solution locally. In each subdomain, a subdomain problem is solved independently and, more importantly, in a much lower dimensional random space. In a postprocesing stage, accurate samples of the original stochastic problems are obtained from the samples of the local solutions by enforcing the correct stochastic structure of the random inputs and the coupling conditions at the interfaces of the subdomains. Overall, the method is able to solve stochastic PDEs in very large dimensions by solving a collection of low dimensional local problems and can be highly efficient. In our paper we present the general mathematical framework of the methodology and use numerical examples to demonstrate the properties of the method.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2011

Linear-optics realization of channels for single-photon multimode qudits

We propose and theoretically study a method for the stochastic realization of arbitrary quantum channels on multimode single-photon qudits. In order for our method to be undemanding in its implementation, we restrict our analysis to linear-optical techniques, vacuum ancillary states, and nonadaptive schemes, but we allow for random switching between different optical networks. With our method it is possible to deterministically implement random-unitary channels and to stochastically implement general channels. We provide an expression for the optimal probability of success of our scheme and calculate this quantity for specific examples such as the qubit amplitude-damping channel. The success probability is shown to be related to the entanglement properties of the Choi-Jamiolkowski state isomorphic to the channel.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2012

Oral Mucositis Prevention By Low-Level Laser Therapy in Head-and-Neck Cancer Patients Undergoing Concurrent Chemoradiotherapy: A Phase III Randomized Study

Purpose: Oral mucositis is a major complication of concurrent chemoradiotherapy (CRT) in head-and-neck cancer patients. Low-level laser (LLL) therapy is a promising preventive therapy. We aimed to evaluate the efficacy of LLL therapy to decrease severe oral mucositis and its effect on RT interruptions. Methods and Materials: In the present randomized, double-blind, Phase III study, patients received either gallium-aluminum-arsenide LLL therapy 2.5 J/cm{sup 2} or placebo laser, before each radiation fraction. Eligible patients had to have been diagnosed with squamous cell carcinoma or undifferentiated carcinoma of the oral cavity, pharynx, larynx, or metastases to the neck with an unknown primary site. They were treated with adjuvant or definitive CRT, consisting of conventional RT 60-70 Gy (range, 1.8-2.0 Gy/d, 5 times/wk) and concurrent cisplatin. The primary endpoints were the oral mucositis severity in Weeks 2, 4, and 6 and the number of RT interruptions because of mucositis. The secondary endpoints included patient-reported pain scores. To detect a decrease in the incidence of Grade 3 or 4 oral mucositis from 80% to 50%, we planned to enroll 74 patients. Results: A total of 75 patients were included, and 37 patients received preventive LLL therapy. The mean delivered radiation dose was greater in the patients treated with LLL (69.4 vs. 67.9 Gy, p = .03). During CRT, the number of patients diagnosed with Grade 3 or 4 oral mucositis treated with LLL vs. placebo was 4 vs. 5 (Week 2, p = 1.0), 4 vs. 12 (Week 4, p = .08), and 8 vs. 9 (Week 6, p = 1.0), respectively. More of the patients treated with placebo had RT interruptions because of mucositis (6 vs. 0, p = .02). No difference was detected between the treatment arms in the incidence of severe pain. Conclusions: LLL therapy was not effective in reducing severe oral mucositis, although a marginal benefit could not be excluded. It reduced RT interruptions in these head-and-neck cancer patients, which might translate into improved CRT efficacy.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2025

Denoising Autoencoder for Reconstructing Sensor Observation Data and Predicting Evapotranspiration: Noisy and Missing Values Repair and Uncertainty Quantification

Abstract Machine learning (ML) methods applied in scientific research often deal with interrelated features in high‐dimensional data. Reducing data noise and redundancy is needed to increase prediction accuracy and efficiency especially when dealing with data from field sensors. We explored an unsupervised learning method, the denoising autoencoder (DAE), to extract the underlying data structure from noisy raw data in the context of predicting hydrologic quantities from multiple field sensors. These sensors have intrinsic instrumental noise and occasional malfunctions that cause missing values. Our DAE neural network reconstructed meteorological sensor data containing noise and missing values to predict evapotranspiration in a mountainous watershed. The DAE reconstructed the sensor variables with a mean coefficient of determination value of 0.77 across 15 dimensions representing individual sensors. It reduced variance and bias uncertainties compared to a classical autoencoder model. The reconstruction quality varied across dimensions depending on their cross‐correlation and alignment with the underlying data structure. Uncertainties arising from the model structure were overall higher than those resulting from data corruption. We attached the DAE structure to a downstream ET‐prediction neural network in three formats and achieved reasonably accurate ET predictions . The use of the DAE notably reduced variance uncertainty in ET prediction. However, excessive variance reduction may be accompanied by an increase in bias due to the intrinsic bias‐variance tradeoff. Our method of evaluating and reducing uncertainties in aggregated data from different sources can be used to improve predictive models, process understanding, and uncertainty quantification for better water resource management. Plain Language Summary We present a machine learning method, namely the denoising autoencoder, which reduces the effects of data noise and missing values typically present in scientific data sets collected through sensor measurements. This method selects the most relevant information from noisy raw data collected by the instruments and fills in missing values. To demonstrate the effectiveness of our method, we applied it to predict evapotranspiration, a hydrologic variable that represents the water moved from the land surface to the atmosphere through a combination of evaporation and plant water use (transpiration). We also used a random sampling technique (the Monte Carlo method) to compare the uncertainty in the predictions when using the raw and noisy data versus the reconstructed data. The denoising process produced more accurate predictions of evapotranspiration with less uncertainty. Improved predictions of evapotranspiration can lead to a better understanding and accounting of water budgets. This ML approach is broadly suitable for a wide variety of applications that involve noisy sensor data with missing values. Key Points We used a denoising autoencoder (DAE) neural network to reduce noise in meteorological and soil sensor observations by on average We used Monte Carlo sampling to estimate the bias and variance of all model outputs, including uncertainty sources from data and the model We attached the DAE component to a downstream neural network to predict ET with the variance reduced by , compared to that without the DAE

denoising autoencoder↗
DOE OSTI2015

Verifying and Validating Simulation Models

This presentation is a high-level discussion of the Verification and Validation (V&V) of computational models. Definitions of V&V are given to emphasize that “validation” is never performed in a vacuum; it accounts, instead, for the current state-of-knowledge in the discipline considered. In particular comparisons between physical measurements and numerical predictions should account for their respective sources of uncertainty. The differences between error (bias), aleatoric uncertainty (randomness) and epistemic uncertainty (ignorance, lack-of- knowledge) are briefly discussed. Four types of uncertainty in physics and engineering are discussed: 1) experimental variability, 2) variability and randomness, 3) numerical uncertainty and 4) model-form uncertainty. Statistical sampling methods are available to propagate, and analyze, variability and randomness. Numerical uncertainty originates from the truncation error introduced by the discretization of partial differential equations in time and space. Model-form uncertainty is introduced by assumptions often formulated to render a complex problem more tractable and amenable to modeling and simulation. The discussion concludes with high-level guidance to assess the “credibility” of numerical simulations, which stems from the level of rigor with which these various sources of uncertainty are assessed and quantified.

Mathematics & Computing(97)↗