Search NASA⌕ Search

SEARCH · Search NASA

Results for “Bayesian method”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,297 records · Page 72

Applications of Ensemble-based Data Assimilation Techniques for Aquifer Characterization using Tracer Data at Hanford 300 Area

Subsurface aquifer characterization often involves high parameter dimensionality and requires tremendous computational resources if employing a full Bayesian approach. Ensemble-based data assimilation techniques, including filtering and smoothing, are computationally efficient alternatives. Despite the increasing number of applications of ensemble-based methods in assimilating flow and transport related data for subsurface aquifer charaterization, most are limited to either synthetic studies or two-dimensional problems. In this study, we applied ensemble-based techniques for assimilating field tracer experimental data obtained from the Integrated Field Research Challenge (IFRC) site at the Hanford 300 Area. The forward problem was simulated using the massively-parallel three-dimensional flow and transport code PFLOTRAN to effectively deal with the highly transient flow boundary conditions at the site and to meet the computational demands of ensemble-based methods. This study demonstrates the effectiveness of ensemble-based methods for characterizing a heterogeneous aquifer by sequentially assimilating multiple types of data. The necessity of employing high performance computing is shown to enable increasingly mechanistic non-linear forward simulations to be performed within the data assimilation framework for a complex system with reasonable turnaround time.

data assimilation, ensemble kalman filter, inverse↗

Quantifying uncertainty for deep learning based forecasting and flow-reconstruction using neural architecture search ensembles

Classical problems in computational physics such as data-driven forecasting and signal reconstruction from sparse sensors have recently seen an explosion in deep neural network (DNN) based algorithmic approaches. However, most DNN models do not provide uncertainty estimates, which are crucial for establishing the trustworthiness of these techniques in downstream decision making tasks and scenarios. In recent years, ensemble-based methods have achieved significant success for the uncertainty quantification in DNNs on a number of benchmark problems. However, their performance on real-world applications remains under-explored. In this work, we present an automated approach to DNN discovery and demonstrate how this may also be utilized for ensemble-based uncertainty quantification. Specifically, we propose the use of a scalable neural and hyperparameter architecture search for discovering an ensemble of DNN models for complex dynamical systems. We highlight how the proposed method not only discovers high-performing neural network ensembles for our tasks, but also quantifies uncertainty seamlessly. This is achieved by using genetic algorithms and Bayesian optimization for sampling the search space of neural network architectures and hyperparameters. Subsequently, a model selection approach is used to identify candidate models for an ensemble set construction. Afterwards, a variance decomposition approach is used to estimate the uncertainty of the predictions from the ensemble. We demonstrate the feasibility of this framework for two tasks — forecasting from historical data and flow reconstruction from sparse sensors for the sea-surface temperature. In conclusion, we demonstrate superior performance from the ensemble in contrast with individual high-performing models and other benchmarks.

Deep ensembles↗

Statistical Uncertainty in Paleoclimate Proxy Reconstructions

A quantitative analysis of any environment older than the instrumental record relies on proxies. Uncertainties associated with proxy reconstructions are often underestimated, which can lead to artificial conflict between different proxies, and between data and models. In this paper, using ordinary least squares linear regression as a common example, we describe a simple, robust and generalizable method for quantifying uncertainty in proxy reconstructions. We highlight the primary controls on the magnitude of uncertainty, and compare this simple estimate to equivalent estimates from Bayesian, nonparametric and fiducial statistical frameworks. We discuss when it may be possible to reduce uncertainties, and conclude that the unexplained variance in the calibration must always feature in the uncertainty in the reconstruction. This directs future research toward explaining as much of the variance in the calibration data as possible. We also advocate for a “data-forward” approach, that clearly decouples the presentation of proxy data from plausible environmental inferences.

58 GEOSCIENCES↗

Maximally local two-nucleon interactions at N 3 LO in Δ -less chiral effective field theory

We present new maximally local two-nucleon interactions derived in Δ -less chiral effective field theory up to next-to-next-to-next-to-leading order ( N 3 LO ), which include all contact and pion-exchange contributions to the nuclear Hamiltonian up to this order. Our interactions are fit to nucleon-nucleon phase shifts using a Bayesian statistical approach, and explore a wide cutoff range from 0.6 – 0.9 fm ( ≈ 660 – 440 MeV ). These interactions can be straightforwardly employed in quantum Monte Carlo methods, such as the auxiliary field diffusion Monte Carlo method. Together with local three-nucleon forces, calculations with these new interactions will provide improved benchmarks for the structure of atomic nuclei and serve as crucial input to analyses of astrophysical phenomena of neutron stars, such as binary neutron-star mergers. Published by the American Physical Society 2024

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Integrated Framework of Multisource Data Fusion for Outage Location in Looped Distribution Systems

Accurate outage location is essential for expediting post-outage power restoration, minimizing outage duration, and enhancing the resilience of distribution networks. With the advent of advanced metering infrastructure, data-driven outage location methods have significantly advanced beyond traditional approaches that rely on manual inspections. However, existing methods still face critical challenges, like reliance on single-source data, limited ability to handle partially observable systems or difficulties with loop networks. To the best of our knowledge, no single approach has comprehensively addressed all of these challenges at once. To this end, this paper proposes a comprehensive multisource data fusion framework for outage locations via probabilistic graph networks. The framework consists of three key phases. First, a novel method for reconstituting distribution networks with loops is developed, transforming looped networks into multiple radial subnetworks that retain all outage causalities of the original network. Second, Bayesian network (BN) models are established for each subnetwork, integrating multiple data sources and network structures. Finally, a joint Gibbs sampling mechanism, featuring forward and backward information flow, is designed to merge data from separate BN models and maximize the utilization of limited evidence, ensuring accurate outage location identification. In conclusion, the framework was validated on two modified public test systems, and comparative studies confirmed its effectiveness.

24 POWER TRANSMISSION AND DISTRIBUTION↗

A Bayesian approach to real-time 3D tumor localization via monoscopic x-ray imaging during treatment delivery

Purpose: Monoscopic x-ray imaging with on-board kV devices is an attractive approach for real-time image guidance in modern radiation therapy such as VMAT or IMRT, but it falls short in providing reliable information along the direction of imaging x-ray. By effectively taking consideration of projection data at prior times and/or angles through a Bayesian formalism, the authors develop an algorithm for real-time and full 3D tumor localization with a single x-ray imager during treatment delivery. Methods: First, a prior probability density function is constructed using the 2D tumor locations on the projection images acquired during patient setup. Whenever an x-ray image is acquired during the treatment delivery, the corresponding 2D tumor location on the imager is used to update the likelihood function. The unresolved third dimension is obtained by maximizing the posterior probability distribution. The algorithm can also be used in a retrospective fashion when all the projection images during the treatment delivery are used for 3D localization purposes. The algorithm does not involve complex optimization of any model parameter and therefore can be used in a ''plug-and-play'' fashion. The authors validated the algorithm using (1) simulated 3D linear and elliptic motion and (2) 3D tumor motion trajectories of a lung and a pancreas patient reproduced by a physical phantom. Continuous kV images were acquired over a full gantry rotation with the Varian TrueBeam on-board imaging system. Three scenarios were considered: fluoroscopic setup, cone beam CT setup, and retrospective analysis. Results: For the simulation study, the RMS 3D localization error is 1.2 and 2.4 mm for the linear and elliptic motions, respectively. For the phantom experiments, the 3D localization error is < 1 mm on average and < 1.5 mm at 95th percentile in the lung and pancreas cases for all three scenarios. The difference in 3D localization error for different scenarios is small and is not statistically significant. Conclusions: The proposed algorithm eliminates the need for any population based model parameters in monoscopic image guided radiotherapy and allows accurate and real-time 3D tumor localization on current standard LINACs with a single x-ray imager.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Q-BEEP: Quantum Bayesian Error Mitigation Employing Poisson Modeling over the Hamming Spectrum

Quantum computing technology has grown rapidly in recent years, with new technologies being explored, error rates being reduced, and quantum processor’s qubit capacity growing. However, near-term quantum algorithms are still unable to be induced without compounding consequential levels of noise, leading to non-trivial erroneous results. Quantum Error Correction (in-situ error mitigation) and Quantum Error Mitigation (post-induction error mitigation) are promising fields of research within the quantum algorithm scene, aiming to alleviate quantum errors, increasing the overall fidelity and hence the overall quality of circuit induction. Earlier this year, a pioneering work, namely HAMMER, published in ASPLOS-22 demonstrated the existence of a latent structure regarding post-circuit induction errors when mapping to the Hamming spectrum. However, they intuitively assumed that errors occur in local clusters, and that at higher average Hamming distances this structure falls away. In this work, we show that such a correlation structure is not only local but extends certain non-local clustering patterns which can be precisely described by a Poisson distribution model taking the input circuit, the device run time status (i.e., calibration statistics) and qubit topology into consideration. Using this quantum error characterizing model, we developed an iterative algorithm over the generated Bayesian network state-graph for post-induction error mitigation. Thanks to more precise modeling of the error distribution latent structure and the new iterative method, our Q-Beep approach provides state of the art performance and can boost circuit execution fidelity by up to 234.6% on Bernstein-Vazirani circuits and on average 71.0% on QAOA solution quality, using 16 practical IBMQ quantum processors. For other benchmarks such as those in QASMBench, the fidelity improvement is up to 17.8%. Q-Beep is a light-weight post-processing technique that can be performed offline and remotely, making it a useful tool for quantum vendors to integrate and provide more reliable circuit induction results.

Stein, Samuel A.↗

Generalized Bayesian Framework for Evaluation of Integral Benchmark Experiments

A recently published generalized Bayesian optimization framework has provided a way to retract any or all of the three common assumptions underlying the conventional Generalized Linear Least Squares (GLLS) optimization method based on the concepts introduced in Ref. [2]. These assumptions are: 1. Perfection: The model used for data evaluation and the prior probability distribution function (PDF) of generalized data are perfect. 2. Normality: The prior and posterior PDF are normal. 3. Linearity: The model is linear. In this work we outline how the framework in [1] could be directly adopted for improved evaluation of nuclear criticality integral benchmark experiments (IBEs) by: 1. Removing the first assumption alone by utilizing the concept of imperfections introduced in [1] to enable evaluation in the presence of discrepancies between the data and model or of missing covariance information by a GLLS method that will be seen as a generalization of the conventional GLLS method employed by the TSURFER code, and by 2. Removing the remaining two assumptions by implementing a Markov Chain Monte Carlo method for computation of the posterior PDF in the SAMPLER code, where TSURFER and SAMPLER are the uncertainty quantification (UQ) codes for IBEs in the SCALE code system based on the GLLS and the stochastic method, respectively. The graphic in Figure 1 categorizes the methods discussed in terms of the assumptions that they employ to determine posterior PDFs.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Generalized Bayesian Framework for Evaluation of Integral Benchmark Experiments

A recently published generalized Bayesian optimization framework has provided a way to retract any or all of the three common assumptions underlying the conventional Generalized Linear Least Squares (GLLS) optimization method based on the concepts introduced in reference two. These assumptions are: 1. Perfection: The model used for data evaluation and the prior probability distribution function (PDF) of generalized* data are perfect; 2. Normality: The prior and posterior PDF are normal; and 3. Linearity: The model is linear. In this work we outline how the framework in 1 could be directly adopted for improved evaluation of nuclear criticality integral benchmark experiments (IBEs) by: 1. Removing the first assumption alone by utilizing the concept of imperfections introduced in 1 to enable evaluation in the presence of discrepancies between the data and model or of missing covariance information by a GLLS method that will be seen as a generalization of the conventional GLLS method employed by the TSURFER code, and by 2. Removing the remaining two assumptions by implement- ing a Markov Chain Monte Carlo method for computation of the posterior PDF in the SAMPLER code, where TSURFER and SAMPLER are the uncertainty quantification (UQ) codes for IBEs in the SCALE code system based on the GLLS, and the stochastic method, respectively.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Impact of Uncertainty on Calculations for Recovery from Loss of Offsite Power

Uncertainty, both aleatory and epistemic, can have a significant impact on estimated probabilities of recovering from loss of offsite power within a specified time window, and such probabilities are an input to risk-informed decisions as to the significance of inspection findings in the U.S. Nuclear Regulatory Commission’s Reactor Oversight Process. In particular, the choice of aleatory model for offsite power recovery time can have a significant impact on the estimated nonrecovery probability, especially if epistemic uncertainty regarding parameters in the aleatory model is accounted for properly. In past and current analyses, such uncertainty has largely been ignored. This paper examines the impact of both aleatory and epistemic uncertainty on the results, using modern open-source Bayesian inference software, which implements Markov chain Monte Carlo sampling. It includes examples of time-dependent convolution calculations to show the impact that uncertainty can have on this increasingly frequent type of calculation, also. The results show that the “point estimate” result, which is an input to risk-informed decisions, can easily be uncertain by a factor of 10 if both aleatory and epistemic uncertainties are considered. The paper also illustrates the use of Bayesian model selection criteria to aid in the choice of aleatory model.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Extending Parsimonious Bayesian Inference

Parsimonious Bayesian inference is a theoretical framework for efficient data assimilation that seeks to balance increased consistency between predictions and training data against corresponding increases in model complexity. Within this framework, over-training is understood as optimization that encodes excessive information within model parameters while only achieving small improvements between predictions and training data. This project aims to develop practical methods of limiting excess model information during optimization. One key observation is that practical heuristics for parsimonious learning in high-dimensions must balance expressivity, i.e. the ability of the model to capture diverse predictions with only a few non-zero parameters, against discoverability, i.e. the ability to train the model with gradient-based optimization and drive parameters to low information states. As such, we developed logical activation functions that are able to adaptively approximate arbitrary truth tables that define Boolean logic operations within a probabilistic framework. These functions have demonstrated the ability to learn exclusive disjunction (XOR) and conditioned disjunction (if [condition] then [result_if_true] else [result_if_false]) within a single layer of a neural network. To efficiently exploit these activation functions to drive parsimonious learning required several other advances within the domain of variational inference. The most efficient form of complexity suppression is structured sparsification, driving most model parameters to zero while achieving the structural coherence among nonzeros needed for bandwidth reduction. Such models are not only far more efficient at suppressing information-theoretic complexity, they also reduce the other forms of complexity (computations, communication, storage, and the number of dependencies needed to evaluate predictions). Aiming to support enhanced sparsification, this project examined new approaches to high-dimensional variational inference that allow us to calibrate and control parameter uncertainty during optimization. By identifying which parameters can sustain sparsifying perturbations with little impact on prediction quality, we can develop better pruning strategies by framing them as approximate Bayesian inference. These advances also open paths to mitigate concerns with deploying advanced learning methods in resource-constrained environments, such as running models on power-limited or communication-limited devices.

97 MATHEMATICS AND COMPUTING↗

Featureless adaptive optimization accelerates functional electronic materials design

Electronic materials that exhibit phase transitions between metastable states (e.g., metal-insulator transition materials with abrupt electrical resistivity transformations) are challenging to decode. For these materials, conventional machine learning methods display limited predictive capability due to data scarcity and the absence of features that impede model training. In this article, we demonstrate a discovery strategy based on multi-objective Bayesian optimization to directly circumvent these bottlenecks by utilizing latent variable Gaussian processes combined with high-fidelity electronic structure calculations for validation in the chalcogenide lacunar spinel family. We directly and simultaneously learn phase stability and bandgap tunability from chemical composition alone to efficiently discover all superior compositions on the design Pareto front. Previously unidentified electronic transitions also emerge from our featureless adaptive optimization engine. Our methodology readily generalizes to optimization of multiple properties, enabling co-design of complex multifunctional materials, especially where prior data is sparse.

36 MATERIALS SCIENCE↗

Value-of-Information Analysis of a Fracture Prediction Method

A core-based fracture prediction method is used to illustrate a Value of Information (VOI) decision-analysis protocol to inform completion decisions in tight-gas sandstones. The ratio of late host-rock cement to available pore volume or degradation index uses petrographic observations of cement distributions in core (including sidewall cores) to predict whether nearby but unsampled fractures (widths > 0.5 to 1 mm) are sealed (non-conductive) or open (conductive). Measurements from four sandstone plays suggest that the index correctly predicts open vs. sealed fractures with an accuracy in excess of 80%. The value added is calculated using Bayesian inference in which accuracy of the index serves as the likelihood of the prior distribution of open fractures to assess the posterior probability that data represent a useful predictor of producibility. VOI of the prediction method is more than three times the cost to acquire the data. VOI is most sensitive to play-specific geologic and cost parameters including cost to drill, expected revenue from a successful well, cost of completion, cost of acquiring data for the index, and fracture probability distributions. The approach provides a way to value acquiring fracture data and points to a need for zone-specific production data in tight-gas sandstones.

02 PETROLEUM↗

An intergalactic medium temperature from a giant radio galaxy

ABSTRACT The warm–hot intergalactic medium (warm–hot IGM, or WHIM) pervades the filaments of the Cosmic Web and harbours half of the Universe’s baryons. The WHIM’s thermodynamic properties are notoriously hard to measure. Here we estimate a galaxy group–WHIM boundary temperature using a new method. In particular, we use a radio image of the giant radio galaxy (giant RG, or GRG) created by NGC 6185, a massive nearby spiral. We analyse this extraordinary object with a Bayesian 3D lobe model and deduce an equipartition pressure Peq = 6 × 10−16 Pa – among the lowest found in RGs yet. Using an X-ray-based statistical conversion for Fanaroff–Riley II RGs, we find a true lobe pressure $P = 1.5\substack{+1.7\\ -0.4}\times 10^{-15}\ \mathrm{Pa}$. Cosmic Web reconstructions, group catalogues, and MHD simulations furthermore imply an Mpc–scale IGM density $1 + \delta _\mathrm{IGM} = 40\substack{+30\\ -10}$. The buoyantly rising lobes are crushed by the IGM at their inner side, where an approximate balance between IGM and lobe pressure occurs: PIGM ≈ P. The ideal gas law then suggests an IGM temperature $T_\mathrm{IGM} = 11\substack{+12\\-5} \times 10^6\ \mathrm{K}$, or $k_\mathrm{B}T_\mathrm{IGM} = 0.9\substack{+1.0\\-0.4}\ \mathrm{keV}$, at the virial radius – consistent with X-ray-derived temperatures of similarly massive groups. Interestingly, the method is not performing at its limit: in principle, estimates TIGM ∼ 4 × 106 K are already possible – rivalling the lowest X-ray measurements available. The technique’s future scope extends from galaxy group outskirts to the WHIM. In conclusion, we demonstrate that observations of GRGs in Cosmic Web filaments are finally sensitive enough to probe the thermodynamics of galaxy groups and beyond.

79 ASTRONOMY AND ASTROPHYSICS↗

The Attraction Indian Buffet Distribution

We propose the attraction Indian buffet distribution (AIBD), a distribution for binary feature matrices influenced by pairwise similarity information. Binary feature matrices are used in Bayesian models to uncover latent variables (i.e., features) that explain observed data. The Indian buffet process (IBP) is a popular exchangeable prior distribution for latent feature matrices. In the presence of additional information, however, the exchangeability assumption is not reasonable or desirable. The AIBD can incorporate pairwise similarity information, yet it preserves many properties of the IBP, including the distribution of the total number of features. Thus, much of the interpretation and intuition that one has for the IBP directly carries over to the AIBD. A temperature parameter controls the degree to which the similarity information affects feature-sharing between observations. Unlike other nonexchangeable distributions for feature allocations, the probability mass function of the AIBD has a tractable normalizing constant, making posterior inference on hyperparameters straight-forward using standard MCMC methods. A novel posterior sampling algorithm is proposed for the IBP and the AIBD. We demonstrate the feasibility of the AIBD as a prior distribution in feature allocation models and compare the performance of competing methods in simulations and an application.

97 MATHEMATICS AND COMPUTING↗

Deep Sigma Point Processes-Assisted Chance-Constrained Power System Transient Stability Preventive Control

Here this paper proposes a deep sigma point processes (DSPP)-assisted chance-constrained power system transient stability preventive control method to deal with uncertain renewable energy and loads-induced stability risk. The traditional transient stability-constrained preventive control is reformulated as a chance-constrained optimization problem. To deal with the computational bottleneck of the time-domain simulation-based probabilistic transient stability assessment, the DSPP is developed. DSPP is a parametric Bayesian approach that allows us to predict system transient stability with high computational efficiency while accurately quantifying the confidence intervals of the predictions that can be used to inform system instability risk. To this end, with a given preset confidence probability, we embed DSPP into the primal dual interior point method to help solve the chance-constrained preventive control problem, where the corresponding Jacobian and Hessian matrices are derived. Comparison results with other existing methods show that the proposed method can significantly speed up preventive control while maintaining high accuracy and convergence

97 MATHEMATICS AND COMPUTING↗

Hierarchical calibration and validation for modeling bench-scale solvent-based carbon capture. Part 1: Non-reactive physical mass transfer across the wetted wall column

A hierarchical model calibration and validation are proposed here for quantifying the confidence level of mass transfer prediction using a computational fluid dynamics (CFD) model, where the solvent-based carbon dioxide (CO 2 ) capture is simulated and the predicted results are compared to the corresponding bench-scale experimental data. Two unit problems with an increasing level of complexity are proposed to break down the complex physical/chemical processes of solvent-based CO 2 capture into relatively simpler problems to separate the effects of physical transport and chemical reaction. This paper focuses on the calibration and validation of the first unit problem, i.e., the CO 2 mass transfer across a falling ethanolamine (MEA) film in the absence of chemical reaction. This problem is investigated both experimentally and numerically using non-reactive nitrous oxide (N 2 O) as a surrogate for CO 2 . To capture motion of the gas-liquid interface, a volume of fluid method is employed with a one-fluid formulation to compute the mass transfer between the two phases. Parallel bench-scale experiments are designed and conducted to validate and calibrate the CFD models using a general Bayesian calibration approach. Two important transport parameters, Henry's constant and gas diffusivity, are calibrated to produce the posterior distributions, which will be used as input for the second unit problem to address the chemical absorption of CO 2 across the MEA falling film, where both mass transfer and chemical reaction are involved. Finally, mass transfer coefficients predicted by CFD are also compared with those predicted by the traditional/empirical correlations under both steady and wavy falling film conditions.

54 ENVIRONMENTAL SCIENCES↗

Evaluation of Neutron-induced Cross Sections and their Related Covariances with Physical Constraints

Nuclear data, along with numerical methods and the associated calculation schemes, continue to play a key role in reactor design, reactor core operating parameters calculations, fuel cycle management and criticality safety calculations. Due to the intensive use of Monte-Carlo calculations reducing numerical biases, the final accuracy of neutronic calculations increasingly depends on the quality of nuclear data used. This paper gives a broad picture of all ingredients treated by nuclear data evaluators during their analyses. After giving an introduction to nuclear data evaluation, we present implications of using the Bayesian inference to obtain evaluated cross sections and related uncertainties. In particular, a focus is made on systematic uncertainties appearing in the analysis of differential measurements as well as advantages and drawbacks one may encounter by analyzing integral experiments. The evaluation work is in general done independently in the resonance and in the continuum energy ranges giving rise to inconsistencies in evaluated files. For future evaluations on the whole energy range, we call attention to two innovative methods used to analyze several nuclear reaction models and impose constraints. Finally, we discuss suggestions for possible improvements in the evaluation process to master the quantification of uncertainties. These are associated with experiments (microscopic and integral), nuclear reaction theories and the Bayesian inference.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗