Search NASA⌕ Search

SEARCH · Search NASA

Results for “Bayesian model selection”

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 307 records · Page 17

Bayesian-information-gap decision theory with an application to CO 2 sequestration

Decisions related to subsurface engineering problems such as groundwater management, fossil fuel production, and geologic carbon sequestration are frequently challenging because of an overabundance of uncertainties (related to conceptualizations, parameters, observations, etc.). Because of the importance of these problems to agriculture, energy, and the climate (respectively), good decisions that are scientifically defensible must be made despite the uncertainties. We describe a general approach to making decisions for challenging problems such as these in the presence of severe uncertainties that combines probabilistic and non-probabilistic methods. The approach uses Bayesian sampling to assess parametric uncertainty and Information-Gap Decision Theory (IGDT) to address model inadequacy. The combined approach also resolves an issue that frequently arises when applying Bayesian methods to real-world engineering problems related to the enumeration of possible outcomes. In the case of zero non-probabilistic uncertainty, the method reduces to a Bayesian method. Lastly, to illustrate the approach, we apply it to a site-selection decision for geologic CO 2 sequestration.

58 GEOSCIENCES↗

Gaussian processes with built-in dimensionality reduction: Applications to high-dimensional uncertainty propagation

Uncertainty quantification (UQ) tasks, such as model calibration, uncertainty propagation, and optimization under uncertainty, typically require several thousand evaluations of the underlying computer codes. To cope with the cost of simulations, one replaces the real response surface with a cheap surrogate based, e.g., on polynomial chaos expansions, neural networks, support vector machines, or Gaussian processes (GP). However, the number of simulations required to learn a generic multivariate response grows exponentially as the input dimension increases. This curse of dimensionality can only be addressed, if the response exhibits some special structure that can be discovered and exploited. A wide range of physical responses exhibit a special structure known as an active subspace (AS). An AS is a linear manifold of the stochastic space characterized by maximal response variation. The idea is that one should first identify this low dimensional manifold, project the high-dimensional input onto it, and then link the projection to the output. If the dimensionality of the AS is low enough, then learning the link function is a much easier problem than the original problem of learning a high-dimensional function. The classic approach to discovering the AS requires gradient information, a fact that severely limits its applicability. Furthermore, and partly because of its reliance to gradients, it is not able to handle noisy observations. The latter is an essential trait if one wants to be able to propagate uncertainty through stochastic simulators, e.g., through molecular dynamics codes. In this work, we develop a probabilistic version of AS which is gradient-free and robust to observational noise. Our approach relies on a novel Gaussian process regression with built-in dimensionality reduction. In particular, the AS is represented as an orthogonal projection matrix that serves as yet another covariance function hyper-parameter to be estimated from the data. To train the model, we design a two-step maximum likelihood optimization procedure that ensures the orthogonality of the projection matrix by exploiting recent results on the Stiefel manifold, i.e., the manifold of matrices with orthogonal columns. The additional benefit of our probabilistic formulation, is that it allows us to select the dimensionality of the AS via the Bayesian information criterion. We validate our approach by showing that it can discover the right AS in synthetic examples without gradient information using both noiseless and noisy observations. We demonstrate that our method is able to discover the same AS as the classical approach in a challenging one-hundred-dimensional problem involving an elliptic stochastic partial differential equation with random conductivity. Finally, we use our approach to study the effect of geometric and material uncertainties in the propagation of solitary waves in a one dimensional granular system.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Technical Report: Algorithm and Implementation for Quasispecies Abundance Inference with Confidence Intervals from Metagenomic Sequence Data

This report describes the design and implementation of an algorithm for estimating relative microbial abundances, together with confidence limits, using data from metagenomic DNA sequencing. For the background behind this project and a detailed discussion of our modeling approach for metagenomic data, we refer the reader to our earlier technical report, dated March 4, 2014. Briefly, we described a fully Bayesian generative model for paired-end sequence read data, incorporating the effects of the relative abundances, the distribution of sequence fragment lengths, fragment position bias, sequencing errors and variations between the sampled genomes and the nearest reference genomes. A distinctive feature of our modeling approach is the use of a Chinese restaurant process (CRP) to describe the selection of genomes to be sampled, and thus the relative abundances. The CRP component is desirable for fitting abundances to reads that may map ambiguously to multiple targets, because it naturally leads to sparse solutions that select the best representative from each set of nearly equivalent genomes.

59 BASIC BIOLOGICAL SCIENCES↗

Relative effect of anthropogenic warming and natural climate variability to changes in Compound drought and heatwaves

Compound drought and heatwave (CDHW) events can be influenced by large scale teleconnections and anthropogenic warming, leading to severe socio-economic impacts across various climate regions. Here, the relative influence of six different teleconnection patterns and anthropogenic global warming on the global CDHW occurrences is quantified systematically using the instrumental data period, 1982–2016. The results from the study suggest a substantial increase in the CDHW events (1–5 events per year) across various parts of the globe at the beginning of 21st century (2000–2016). A Bayesian approach is implemented to identify the most vulnerable climate regions based on the degree of susceptibility of heatwaves (DSHW) towards drought. As such, top ten most vulnerable regions are selected based on the DSHW magnitude, and a partial correlation analysis is performed to select the natural and anthropogenic drivers of CDHW in those regions, separately. A logistic regression model is then used to determine significant changes in the odds of CDHW due to changes in the selected drivers that suggest a significantly positive, and multiplicative effect of anthropogenic global warming in the top ten most vulnerable climate regions. Finally, the same logistic regression model, integrated with an analytical framework, is applied to determine the relative influence of anthropogenic global warming on the changes in odds of CDHW for the future, 1.5 °C and 2 °C warming limits. Finally, the results suggest that relative to the 2 °C global warming, constraining to the 1.5 °C global warming limit may conduce about 17-fold reduction in the odds of CDHW in the most vulnerable climate region, East Asia, 5–8-fold reduction in Western North America, Northern Australia, Central North America, Central Europe, South Asia, and the Mediterranean region, and 3–4-fold reduction in Northeastern Brazil, Eastern North America, and West Asia.

54 ENVIRONMENTAL SCIENCES↗

Multi-objective Bayesian optimization of ferroelectric materials with interfacial control for memory and energy storage applications

Optimization of materials’ performance for specific applications often requires balancing multiple aspects of materials’ functionality. Even for the cases where a generative physical model of material behavior is known and reliable, this often requires search over multidimensional function space to identify low-dimensional manifold corresponding to the required Pareto front. In this work, we introduce the multi-objective Bayesian optimization (MOBO) workflow for the ferroelectric/antiferroelectric performance optimization for memory and energy storage applications based on the numerical solution of the Ginzburg–Landau equation with electrochemical or semiconducting boundary conditions. MOBO is a low computational cost optimization tool for expensive multi-objective functions, where we update posterior surrogate Gaussian process models from prior evaluations and then select future evaluations from maximizing an acquisition function. Using the parameters for a prototype bulk antiferroelectric (PbZrO 3 ), we first develop a physics-driven decision tree of target functions from the loop structures. We further develop a physics-driven MOBO architecture to explore multidimensional parameter space and build Pareto-frontiers by maximizing two target functions jointly—energy storage and loss. This approach allows for rapid initial materials and device parameter selection for a given application and can be further expanded toward the active experiment setting. The associated notebooks provide both the tutorial on MOBO and allow us to reproduce the reported analyses and apply them to other systems (https://github.com/arpanbiswas52/MOBO_AFI_Supplements).

36 MATERIALS SCIENCE↗

The Dark Energy Survey supernova program: cosmological biases from supernova photometric classification

ABSTRACT Cosmological analyses of samples of photometrically identified type Ia supernovae (SNe Ia) depend on understanding the effects of ‘contamination’ from core-collapse and peculiar SN Ia events. We employ a rigorous analysis using the photometric classifier SuperNNova on state-of-the-art simulations of SN samples to determine cosmological biases due to such ‘non-Ia’ contamination in the Dark Energy Survey (DES) 5-yr SN sample. Depending on the non-Ia SN models used in the SuperNNova training and testing samples, contamination ranges from 0.8 to 3.5 per cent, with a classification efficiency of 97.7–99.5 per cent. Using the Bayesian Estimation Applied to Multiple Species (BEAMS) framework and its extension BBC (‘BEAMS with Bias Correction’), we produce a redshift-binned Hubble diagram marginalized over contamination and corrected for selection effects, and use it to constrain the dark energy equation-of-state, w. Assuming a flat universe with Gaussian ΩM prior of 0.311 ± 0.010, we show that biases on w are <0.008 when using SuperNNova, with systematic uncertainties associated with contamination around 10 per cent of the statistical uncertainty on w for the DES-SN sample. An alternative approach of discarding contaminants using outlier rejection techniques (e.g. Chauvenet’s criterion) in place of SuperNNova leads to biases on w that are larger but still modest (0.015–0.03). Finally, we measure biases due to contamination on w0 and wa (assuming a flat universe), and find these to be <0.009 in w0 and <0.108 in wa, 5 to 10 times smaller than the statistical uncertainties for the DES-SN sample.

79 ASTRONOMY AND ASTROPHYSICS↗

A BAYESIAN MONTE CARLO ANALYSIS OF THE M-{sigma} RELATION

We present an analysis of selection biases in the M{sub bh}-{sigma} relation using Monte Carlo simulations including the sphere of influence resolution selection bias and a selection bias in the velocity dispersion distribution. We find that the sphere of influence selection bias has a significant effect on the measured slope of the M{sub bh}-{sigma} relation, modeled as {beta}{sub intrinsic} = -4.69 + 2.22{beta}{sub measured}, where the measured slope is shallower than the model slope in the parameter range of {beta} > 4, with larger corrections for steeper model slopes. Therefore, when the sphere of influence is used as a criterion to exclude unreliable measurements, it also introduces a selection bias that needs to be modeled to restore the intrinsic slope of the relation. We find that the selection effect due to the velocity dispersion distribution of the sample, which might not follow the overall distribution of the population, is not important for slopes of {beta} {approx} 4-6 of a logarithmically linear M{sub bh}-{sigma} relation, which could impact some studies that measure low (e.g., {beta} < 4) slopes. Combining the selection biases in velocity dispersions and the sphere of influence cut, we find that the uncertainty of the slope is larger than the value without modeling these effects and estimate an intrinsic slope of {beta} = 5.28{sup +0.84}{sub -0.55}.

79 ASTRONOMY AND ASTROPHYSICS↗

Bayesian tensorized neural networks with automatic rank selection

Tensor decomposition is an effective approach to compress over-parameterized neural networks and to enable their deployment on resource-constrained hardware platforms. However, directly applying tensor compression in the training process is a challenging task due to the difficulty of choosing a proper tensor rank. In order to address this challenge, this paper proposes a low-rank Bayesian tensorized neural network. Our Bayesian method performs automatic model compression via an adaptive tensor rank determination. We also present approaches for posterior density calculation and maximum a posteriori (MAP) estimation for the end-to-end training of our tensorized neural network. Here, we provide experimental validation on a two-layer fully connected neural network, a 6-layer CNN and a 110-layer residual neural network where our work produces 7.4x to 137x more compact neural networks directly from the training while achieving high prediction accuracy.

Low-rank tensor↗

Optimizing long-term monitoring of radiation air-dose rates after the Fukushima Daiichi Nuclear Power Plant

Radiation air dose rates near the Fukushima Daiichi Nuclear Power Plant (FDNPP) have been steadily decreasing over the past eight years since the release of radioactive elements in March 2011. Currently, the radiation monitoring program is expected to transition to long-term monitoring after most of the remediation activities are completed. The main long-term monitoring objectives are to (1) confirm the continuing reduction of contaminant and hazard levels, (2) provide assurance for the public, (3) accumulate the basic datasets for scientific knowledge and future preparation, and (4) detect changes or anomalies in contaminant mobility (if they occur), or any unexpected processes or events. In this work, we have developed a methodology for optimizing the monitoring locations of radiation air dose-rate monitoring. Our approach consists of three steps in order to determine monitoring locations in a systematic manner: (1) prioritizing the critical locations, such as schools or regulatory requirement locations, (2) diversifying locations that cover the key environmental controls that are known to influence contaminant mobility and distributions, and (3) capturing the heterogeneity of radiation air-dose rates across the domain. Therefore, for the second step, we use a Gaussian mixture model to identify the representative locations among multiple environmental variables, such as elevation and land-cover types. For the third step, we use a Gaussian process model to capture and estimate the heterogeneity of air-dose rates across the domain. Employing an integrated dose-rate map derived from Bayesian geostatistical methods as a reference map, we distribute the monitoring locations in such a way as to capture the heterogeneity of the reference map. Our results have shown that this approach allows us to select monitoring locations in a systematic manner such that the heterogeneity of air dose rates is captured by the minimal number of monitoring locations.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

Estimating time since infection in early homogeneous HIV-1 samples using a poisson model

The occurrence of a genetic bottleneck in HIV sexual or mother-to-infant transmission has been well documented. This results in a majority of new infections being homogeneous, i.e., initiated by a single genetic strain. Early after infection, prior to the onset of the host immune response, the viral population grows exponentially. In this simple setting, an approach for estimating evolutionary and demographic parameters based on comparison of diversity measures is a feasible alternative to the existing Bayesian methods (e.g., BEAST), which are instead based on the simulation of genealogies. We have devised a web tool that analyzes genetic diversity in acutely infected HIV-1 patients by comparing it to a model of neutral growth. More specifically, we consider a homogeneous infection (i.e., initiated by a unique genetic strain) prior to the onset of host-induced selection, where we can assume a random accumulation of mutations. Previously, we have shown that such a model successfully describes about 80% of sexual HIV-1 transmissions provided the samples are drawn early enough in the infection. Violation of the model is an indicator of either heterogeneous infections or the initiation of selection. When the underlying assumptions of our model (homogeneous infection prior to selection and fast exponential growth) are met, we are under a very particular scenario for which we can use a forward approach (instead of backwards in time as provided by coalescent methods). This allows for more computationally efficient methods to derive the time since the most recent common ancestor. Furthermore, the tool performs statistical tests on the Hamming distance frequency distribution, and outputs summary statistics (mean of the best fitting Poisson distribution, goodness of fit p-value, etc). The tool runs within minutes and can readily accommodate the tens of thousands of sequences generated through new ultradeep pyrosequencing technologies. The tool is available on the LANL website.

59 BASIC BIOLOGICAL SCIENCES↗

Game Theoretic Approach for Understanding and Modeling Clinical Pathways (Stable Ischemic Heart Disease)

This research develops a game-theoretic agent model for stable ischemic heart disease (SIHD) clinical decisions, incorporating aspects of Bayesian game theory to model hidden information such as the patient’s true underlying medical condition(s). Preexisting observational medical data contained in the United States Department of Veterans Affairs (VA) Corporate Data Warehouse (CDW) is analyzed to obtain the required game-theoretic parameter values. Various statistical adjustments are performed to correct for bias in the preexisting treatment protocol selections and to obtain game-theoretic parameter values for contrafactual treatment protocol choices. Once the game-theoretic parameter values are determined, clinical pathways can be viewed as being mathematical strategies for playing the game and thus can be studied and evaluated using the techniques of game theory.

59 BASIC BIOLOGICAL SCIENCES↗

Towards a New Supply Chain Cybersecurity Risk Analysis Technique

Supply chain cyber-attacks, such as the SolarWinds Orion attack, are occurring with greater frequency. These attacks compromise a digital device before it is sent to customers, bypassing traditional security controls to remain persistent and undetected in operational environments. While supply chain attacks are prevalent, methods for analyzing the risk of these attacks are currently unavailable. This paper proposes new supply chain cyber-attack difficulty and risk metrics to evaluate the relative risk of an attack throughout the supply chain lifecycle. Difficulty metrics for each stakeholder in a digital device’s supply chain (e.g., hardware manufacturing, firmware development, software development, storage, and distribution entities) are calculated using scores from cybersecurity maturity questionnaires in a Bayesian Network leaky Noisy-MAX model. These difficulty metrics are then used to calculate an overall supply chain cyber-attack risk. Vulnerability and recoverability metrics are also proposed to evaluate the relative stakeholder influence in the attack risk. These proposed relative risk metrics enable continuous supply chain monitoring, provide decision-makers with information necessary for improved supplier selection, and help drive improvements in the cybersecurity posture of the stakeholders in their supply chain.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Machine Learning Guided Synthesis of Flash Graphene

Advances in nanoscience have enabled the synthesis of nanomaterials, such as graphene, from low-value or waste materials through flash Joule heating. Though this capability is promising, the complex and entangled variables that govern nanocrystal formation in the Joule heating process remain poorly understood. In this work, machine learning (ML) models are constructed to explore the factors that drive the transformation of amorphous carbon into graphene nanocrystals during flash Joule heating. An XGBoost regression model of crystallinity achieves an r 2 score of 0.8051 ± 0.054. Feature importance assays and decision trees extracted from these models reveal key considerations in the selection of starting materials and the role of stochastic current fluctuations in flash Joule heating synthesis. Furthermore, partial dependence analyses demonstrate the importance of charge and current density as predictors of crystallinity, implying a progression from reaction-limited to diffusion-limited kinetics as flash Joule heating parameters change. Finally, a practical application of the ML models is shown by using Bayesian meta-learning algorithms to automatically improve bulk crystallinity over many Joule heating reactions. Furthermore, these results illustrate the power of ML as a tool to analyze complex nanomanufacturing processes and enable the synthesis of 2D crystals with desirable properties by flash Joule heating.

01 COAL, LIGNITE, AND PEAT↗

Uncertainty quantification for deep learning in particle accelerator applications

With the advent of increased computational resources and improved algorithms, machine learning-based models are being increasingly applied to complex problems in particle accelerators. However, such data-driven models may provide overly confident predictions with unknown errors and uncertainties. For reliable deployment of machine learning models in high-regret and safety-critical systems such as particle accelerators, estimates of prediction uncertainty are needed along with accurate point predictions. In this investigation, we evaluate Bayesian neural networks (BNN) as an approach that can provide accurate predictions along with reliably quantified uncertainties for particle accelerator problems, and compare their performance with bootstrapped ensembles of neural networks. We select three accelerator setups for this evaluation: a storage ring, a photoinjector, and a linac. The problems span different data volumes and dimensionalities (e.g., scalar predictions as well as image outputs). It is found that BNN provide accurate predictions of the mean along with reliable estimates of predictive uncertainty across the test cases. In this vein, BNN may offer an attractive alternative to deterministic deep learning tools to generate accurate predictions with quantified uncertainties in particle accelerator applications.

43 PARTICLE ACCELERATORS↗

Selecting the selector: Comparison of update rules for discrete global optimization

In this paper, we compare some well-known Bayesian global optimization methods in four distinct regimes, corresponding to high and low levels of measurement noise and to high and low levels of “quenched noise” (which term we use to describe the roughness of the function we are trying to optimize). We isolate the two stages of this optimization in terms of a “regressor,” which fits a model to the data measured so far, and a “selector,” which identifies the next point to be measured. Finally, the focus of this paper is to investigate the choice of selector when the regressor is well matched to the data.

97 MATHEMATICS AND COMPUTING↗

Steve: A Hierarchical Bayesian Model for Supernova Cosmology

Here, we present a new Bayesian hierarchical model (BHM) named Steve for performing Type Ia supernova (SN Ia) cosmology fits. This advances previous works by including an improved treatment of Malmquist bias, accounting for additional sources of systematic uncertainty, and increasing numerical efficiency. Given light-curve fit parameters, redshifts, and host-galaxy masses, we fit Steve simultaneously for parameters describing cosmology, SN Ia populations, and systematic uncertainties. Selection effects are characterized using Monte Carlo simulations. We demonstrate its implementation by fitting realizations of SN Ia data sets where the SN Ia model closely follows that used in Steve. Next, we validate on more realistic SNANA simulations of SN Ia samples from the Dark Energy Survey and low-redshift surveys (DES Collaboration et al. 2018). These simulated data sets contain more than 60,000 SNe Ia, which we use to evaluate biases in the recovery of cosmological parameters, specifically the equation of state of dark energy, w. This is the most rigorous test of a BHM method applied to SN Ia cosmology fitting and reveals small w biases that depend on the simulated SN Ia properties, in particular the intrinsic SN Ia scatter model. This w bias is less than 0.03 on average, less than half the statistical uncertainty on w. These simulation test results are a concern for BHM cosmology fitting applications on large upcoming surveys; therefore, future development will focus on minimizing the sensitivity of Steve to the SN Ia intrinsic scatter model.

79 ASTRONOMY AND ASTROPHYSICS↗

Sparse Bayesian Regression with Integrated Feature Selection for Nuclear Reactor Analysis

High-dimensional, nonlinear function estimation using large datasets is a current area of interest in the machine learning community, and applications may be found throughout the analytical sciences, where ever-growing datasets are making more information available to the analyst. In this paper, we leverage the existing relevance vector machine, a sparse Bayesian version of the well-studied support vector machine, and expand the method to include integrated feature selection and automatic function shaping. These innovations produce an algorithm that is able to distinguish variables that are useful for making predictions of a response from variables that are unrelated or confusing. We test the technology using synthetic data, conduct initial performance studies, and develop a model capable of making position-independent predictions of the coreaveraged burnup using a single specimen drawn randomly from a nuclear reactor core.

Dayman, Ken J↗

How to estimate soil organic carbon stocks of agricultural fields? perspectives using ex-ante evaluation

Estimating soil organic carbon (SOC) stocks of agricultural fields has a range of important applications from development of sustainable management practices to monitoring carbon stocks. There are many estimation strategies with the potential for more reliable estimates of SOC stock and more efficient use of soil sampling and analysis resources, especially by leveraging readily available auxiliary information such as remote sensing. However, concrete guidance for strategy selection is lacking. This study narrows this gap with a comparison of strategies for estimating deep SOC stock (0–60 cm) in a prototypical field. Using high density SOC stock measurements and simulation, we built on past studies by 1) ex-ante evaluating a large number of strategy options, 2) using a Bayesian approach to quantify the uncertainty of the comparison, and 3) considering multiple Bayesian models to assess sensitivity to this modeling choice. We found that, using readily available auxiliary information, both balanced and stratified sampling offer substantial improvements over simple random sampling. The auxiliary information most important for this improvement is a Sentinel-2 SOC index = blue / (green × red), followed by the topographic wetness index. We found that these results are robust to the choice of mapping method, but that there is uncertainty in the magnitude of improvement. Here, we recommend future studies implement this Bayesian approach for simulated ex-ante evaluation of SOC stock estimation strategies across more fields to investigate the generalizability of these findings.

54 ENVIRONMENTAL SCIENCES↗