Search NASA⌕ Search

SEARCH · Search NASA

Results for “Gaussian Process Models”

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 253 records · Page 14

A novel methodology for gamma-ray spectra dataset procurement over varying standoff distances and source activities

The adoption of machine learning approaches for gamma-ray spectroscopy has received considerable attention in the literature. Many studies have investigated the deployment of various algorithm architectures to a specific task. However, little attention has been afforded to the development of the datasets leveraged to train the models. Such training datasets typically span a set of environmental or detector parameters to encompass a problem space of interest to a user. Variations in these measurement parameters will also induce fluctuations in the detector response, including expected pile-up and ground scatter effects. Fundamental to this work is the understanding that 1) the underlying spectral shape varies as the measurement parameters change and 2) the statistical uncertainties associated with two spectra impact their level of similarity. While previous studies attribute some arbitrary discretization to the measurement parameters for the generation of their synthetic training data, this work introduces a principled methodology for efficient spectral-based discretization of a problem space. A signal-to-noise ratio (SNR) respective spectral comparison measure and a Gaussian Process Regression (GPR) model are used to predict the spectral similarity across a range of measurement parameters. This innovative approach effectively showcased its capability by dividing a problem space, ranging from 5 cm to 100 cm standoff distances and 5 μCi–100 μCi of 137 Cs, into three unique combinations of measurement parameters. The findings from this work will aid in creating more robust datasets, which incorporate many possible measurement scenarios, reduce the number of required experimental test set measurements, and possibly enable experimental training data collection for gamma-ray spectroscopy.

data science↗

A 0.8-2.4 μm Transmission spectrum of the hot Jupiter CoRoT-1b

Hot Jupiters with brightness temperatures ≳2000 K can have TiO and VO molecules as gaseous species in their atmospheres. The TiO and VO molecules can potentially induce temperature inversions in hot Jupiter atmospheres and also have an observable signature of large optical to infrared transit depth ratios. Previous transmission spectra of very hot Jupiters have shown a lack of TiO and VO, but only in planets that also appear to lack temperature inversions. We measure the transmission spectrum of CoRoT-1b, a hot Jupiter that was predicted to have a temperature inversion potentially due to significant TiO and VO in its atmosphere. We employ the multi-object spectroscopy method using the SpeX and MORIS instruments on the Infrared Telescope Facility (IRTF) and the Gaussian process method to model red noise. By using a simultaneous reference star on the slit for calibration and a wide slit to minimize slit losses, we achieve transit depth precision of 0.03%-0.09%, comparable to the atmospheric scale height but detect no statistically significant molecular features. We combine our IRTF data with optical CoRoT transmission measurements to search for differences in the optical and near-infrared absorption that would arise from TiO/VO. Our IRTF spectrum and the CoRoT photometry disfavor a TiO/VO-rich spectrum for CoRoT-1b, suggesting that the atmosphere has another absorber that could create a temperature inversion or that the blackbody-like emission from the planet is due to a spectroscopically flat cloud, dust, or haze layer that smoothes out molecular features in both CoRoT-1b's emission and transmission spectra. This system represents the faintest planet hosting star (K = 12.2) with a measured planetary transmission spectrum.

79 ASTRONOMY AND ASTROPHYSICS↗

Solving Stochastic Inverse Problems for Property–Structure Linkages Using Data-Consistent Inversion and Machine Learning

Determining process–structure–property linkages is one of the key objectives in material science, and uncertainty quantification plays a critical role in understanding both process–structure and structure–property linkages. In this work, we seek to learn a distribution of microstructure parameters that are consistent in the sense that the forward propagation of this distribution through a crystal plasticity finite element model matches a target distribution on materials properties. This stochastic inversion formulation infers a distribution of acceptable/consistent microstructures, as opposed to a deterministic solution, which expands the range of feasible designs in a probabilistic manner. Furthermore, to solve this stochastic inverse problem, we employ a recently developed uncertainty quantification framework based on push-forward probability measures, which combines techniques from measure theory and Bayes’ rule to define a unique and numerically stable solution. This approach requires making an initial prediction using an initial guess for the distribution on model inputs and solving a stochastic forward problem. To reduce the computational burden in solving both stochastic forward and stochastic inverse problems, we combine this approach with a machine learning Bayesian regression model based on Gaussian processes and demonstrate the proposed methodology on two representative case studies in structure–property linkages.

36 MATERIALS SCIENCE↗

Scalable and efficient learning from crowds with Gaussian processes

Over the last few years, multiply-annotated data has become a very popular source of information. Online platforms such as Amazon Mechanical Turk have revolutionized the labelling process needed for any classification task, sharing the effort between a number of annotators (instead of the classical single expert). This crowdsourcing approach has introduced new challenging problems, such as handling disagreements on the annotated samples or combining the unknown expertise of the annotators. Probabilistic methods, such as Gaussian Processes (GP), have proven successful to model this new crowdsourcing scenario. However, GPs do not scale up well with the training set size, which makes them prohibitive for medium-to-large datasets (beyond 10K training instances). This constitutes a serious limitation for current real-world applications. In this work, we introduce two scalable and efficient GP-based crowdsourcing methods that allow for processing previously-prohibitive datasets. The first one is an efficient and fast approximation to GP with squared exponential (SE) kernel. The second allows for learning a more flexible kernel at the expense of a heavier training (but still scalable to large datasets). Since the latter is not a GP-SE approximation, it can be also considered as a whole new scalable and efficient crowdsourcing method, useful for any dataset size. Both methods use Fourier features and variational inference, can predict the class of new samples, and estimate the expertise of the involved annotators. A complete experimentation compares them with state-of-the-art probabilistic approaches in synthetic and real crowdsourcing datasets of different sizes. Finally, they stand out as the best performing approach for large scale problems. Moreover, the second method is competitive with the current state-of-the-art for small datasets.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

Hypothesis Learning in Automated Experiment: Application to Combinatorial Materials Libraries

Machine learning is rapidly becoming an integral part of experimental physical discovery via automated and high-throughput synthesis, and active experiments in scattering and electron/probe microscopy. This, in turn, necessitates the development of active learning methods capable of exploring relevant parameter spaces with the smallest number of steps. In this work, an active learning approach based on conavigation of the hypothesis and experimental spaces is introduced. This is realized by combining the structured Gaussian processes containing probabilistic models of the possible system's behaviors (hypotheses) with reinforcement learning policy refinement (discovery). This approach closely resembles classical human-driven physical discovery, when several alternative hypotheses realized via models with adjustable parameters are tested during an experiment. This approach is demonstrated for exploring concentration-induced phase transitions in combinatorial libraries of Sm-doped BiFeO 3 using piezoresponse force microscopy, but it is straightforward to extend it to higher-dimensional parameter spaces and more complex physical problems once the experimental workflow and hypothesis generation are available.

36 MATERIALS SCIENCE↗

Search for the associated production of charm quarks and a Higgs boson decaying into a photon pair with the ATLAS detector

A search for the production of a Higgs boson and one or more charm quarks, in which the Higgs boson decays into a photon pair, is presented. This search uses proton-proton collision data with a centre-of-mass energy of $\sqrt{s}$ = 13 TeV and an integrated luminosity of 140 fb −1 recorded by the ATLAS detector at the Large Hadron Collider. The analysis relies on the identification of charm-quark-containing jets, and adopts an approach based on Gaussian process regression to model the non-resonant di-photon background. The observed (expected, assuming the Standard Model signal) upper limit at the 95% confidence level on the cross-section for producing a Higgs boson and at least one charm-quark-containing jet that passes a fiducial selection is found to be 10.6 pb (8.8 pb). The observed (expected) measured cross-section for this process is 5.3 ± 3.2 pb (2.9 ± 3.1 pb).

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Polymers for Extreme Conditions Designed Using Syntax-Directed Variational Autoencoders

We report the design/discovery of new materials is highly nontrivial owing to the near-infinite possibilities of material candidates and multiple required property/performance objectives. Thus, machine learning tools are now commonly employed to virtually screen material candidates with desired properties by learning a theoretical mapping from material-to-property space, referred to as the forward problem. However, this approach is inefficient and severely constrained by the candidates that the human imagination can conceive. Thus, in this work on polymers, we tackle the materials discovery challenge by solving the inverse problem: directly generating candidates that satisfy desired property/performance objectives. We utilize syntax-directed variational autoencoders (VAE) in tandem with Gaussian process regression (GPR) models to discover polymers expected to be robust under three extreme conditions: (1) high temperatures, (2) high electric field, and (3) high temperature and high electric field, useful for critical structural, electrical, and energy storage applications. This approach to learn from and augment) human ingenuity is general and can be extended to discover polymers with other targeted properties and performance measures.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A Machine Learning Method for the Forensics Attribution of Separated Plutonium

Plutonium (Pu) source attribution would be a powerful tool to support nuclear nonproliferation efforts. This capability to find the source of a Pu sample would act as a deterrent to smuggling efforts, and also help regulatory agencies verify declared nuclear activities. Work at Texas A&M University yielded a nuclear forensics methodology, which is capable of determining separated Pu’s reactor of origin, fuel burnup, and the time since irradiation (TSI)—three parameters of interest. The methodology used a set of ten intra-element isotopic ratios found in separated Pu, which was compared to a library of isotopic ratio values produced using neutronics simulations for reactors of interest. By calculating the probability that unknown Pu sample’s isotopic ratio set matched a set in the library, the methodology could predict the three parameters of interest of the sample. One shortcoming of this methodology was an inability to correctly attribute spoofed Pu, where Pu sourced from two different reactors or two different fuel burnup levels are mixed. A new methodology to rectify this vulnerability using machine learning (ML) technique is developed, instead of the maximum likelihood calculation previously used and the results are satisfactory. The ML approach leverages the existing simulated data for training the algorithm, but use them efficiently by only using intra-element isotope ratios that contribute to the attribution one of the three parameters at a time. Previously, all isotope ratios were used to attribute all three parameters together. The new methodology attributes the Pu parameters in three steps, one for each parameter, rather than resolving all of the three parameters simultaneously like the previous maximum likelihood approach. First, a support vector machine classifier with a set of seven isotopic ratios finds the reactor of origin and a set of regression models trained using gaussian process predicts the burnup with a different set of seven isotopes. Finally, TSI is calculated analytically using decay equations. Thus far, the new methodology is capable of attributing pure Pu samples and has been validated using experimental data. The next step will to be augment the classifier training data set with spoofed Pu data.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

RADIAL VELOCITY OBSERVATIONS AND LIGHT CURVE NOISE MODELING CONFIRM THAT KEPLER-91b IS A GIANT PLANET ORBITING A GIANT STAR

Kepler-91b is a rare example of a transiting hot Jupiter around a red giant star, providing the possibility to study the formation and composition of hot Jupiters under different conditions compared to main-sequence stars. However, the planetary nature of Kepler-91b, which was confirmed using phase-curve variations by Lillo-Box et al., was recently called into question based on a re-analysis of Kepler data. We have obtained ground-based radial velocity observations from the Hobby-Eberly Telescope and unambiguously confirm the planetary nature of Kepler-91b by simultaneously modeling the Kepler and radial velocity data. The star exhibits temporally correlated noise due to stellar granulation which we model as a Gaussian Process. We hypothesize that it is this noise component that led previous studies to suspect Kepler-91b to be a false positive. Our work confirms the conclusions presented by Lillo-Box et al. that Kepler-91b is a 0.73 ± 0.13 M {sub Jup} planet orbiting a red giant star.

79 ASTRONOMY AND ASTROPHYSICS↗

Stochastic and a posteriori optimization to mitigate coil manufacturing errors in stellarator design

Abstract It was recently shown in Wechsung et al (2022 Proc. Natl Acad. Sci. USA 119 e2202084119) that there exist electromagnetic coils that generate magnetic fields, which are excellent approximations to quasi-symmetric fields and have very good particle confinement properties. Using a Gaussian process-based model for coil perturbations, we investigate the impact of manufacturing errors on the performance of these coils. We show that even fairly small errors result in noticeable performance degradation. While stochastic optimization yields minor improvements, it is not possible to mitigate these errors significantly. As an alternative to stochastic optimization, we then formulate a new optimization problem for computing optimal adjustments of the coil positions and currents without changing the shapes of the coil. These a-posteriori adjustments are able to reduce the impact of coil errors by an order of magnitude, providing a new perspective for dealing with manufacturing tolerances in stellarator design.

Physics↗

Ensembles of novelty detection classifiers for structural health monitoring using guided waves

Guided wave structural health monitoring uses sparse sensor networks embedded in sophisticated structures for defect detection and characterization. The biggest challenge of those sensor networks is developing robust techniques for reliable damage detection under changing environmental and operating conditions. To address this challenge, we develop a novelty classifier for damage detection based on one class support vector machines. We identify appropriate features for damage detection and introduce a feature aggregation method which quadratically increases the number of available training observations.We adopt a two-level voting scheme by using an ensemble of classifiers and predictions. Each classifier is trained on a different segment of the guided wave signal, and each classifier makes an ensemble of predictions based on a single observation. Using this approach, the classifier can be trained using a small number of baseline signals. We study the performance using monte-carlo simulations of an analytical model and data from impact damage experiments on a glass fiber composite plate.We also demonstrate the classifier performance using two types of baseline signals: fixed and rolling baseline training set. The former requires prior knowledge of baseline signals from all environmental and operating conditions, while the latter does not and leverages the fact that environmental and operating conditions vary slowly over time and can be modeled as a Gaussian process.

Dib, Gerges (ORCID:0000000255004047)↗

Using AI to predict calibration constants for the central drift chamber in GlueX at Jefferson Lab

The AI for Experimental Controls project team at Jefferson Lab has developed an AI system to control and calibrate a large drift chamber system in near-real time. The AI system will monitor environmental and experimental variables to recommend voltage settings that maintain consistent dE/dx gain and optimal resolution throughout the experiment. At present, calibrations are performed after data have been recorded and require a considerable amount of time and attention from experts. The calibrations currently require multiple iterations and depend on accurate tracking information. Our approach uses environmental data, such as atmospheric pressure and gas temperature, and beam conditions, such as the flux of incident particles, as inputs to a Gaussian Process Regression (GPR) model. For the data taken during the GlueX 2020 run period, the GPR is able to predict the existing gain correction factors to within 3.5%. This talk will briefly describe the development, testing, and future plans for this system at Jefferson Lab.

Jeske, Torri↗

Constraining the Milky Way’s ultraviolet-to-infrared SED with Gaussian process regression

ABSTRACT Improving our knowledge of global Milky Way (MW) properties is critical for connecting the detailed measurements only possible from within our Galaxy to our understanding of the broader galaxy population. We here train Gaussian process regression (GPR) models on SDSS (Sloan Digital Sky Survey) galaxies to map from galaxy properties (stellar mass, apparent axial ratio, star formation rate, bulge-to-total ratio, disc scale length, and bar vote fraction) to ultraviolet (UV; GALEX FUV/NUV), optical (SDSS ugriz), and infrared (IR; 2MASS JHKs and WISE W1/W2/W3/W4) fluxes and uncertainties. With these models, we estimate the photometric properties of the MW, resulting in a full UV-to-IR spectral energy distribution (SED) as it would be measured externally, viewed face-on. We confirm that the MW lies in the green valley in optical diagnostic diagrams, but show for the first time that the MW is in the star-forming region in standard UV and IR diagnostics – characteristic of the population of red spiral galaxies. Although our GPR method predicts one band at a time, the resulting MW UV–IR SED is consistent with SEDs of local spirals with characteristics broadly similar to the MW, suggesting that these independent predictions can be combined reliably. Our UV–IR SED will be invaluable for reconstructing the MW’s star formation history using the same tools employed for external galaxies, allowing comparisons of results from in situ measurements to those from the methods used for extragalactic objects.

79 ASTRONOMY AND ASTROPHYSICS↗

Fast baryonic field painting for Sunyaev-Zel’dovich analyses: Transfer function vs hybrid effective field theory

Here, we present two approaches for “painting” baryonic properties relevant to the Sunyaev-Zel’dovich (SZ) effect—optical depth and Compton-y—onto three-dimensional N-body simulations, using the MillenniumTNG suite as a benchmark. The goal of these methods is to produce fast and accurate reconstruction methods to aid future analyses of baryonic feedback using the SZ effect. The first approach employs a Gaussian process emulator to model the SZ quantities via a transfer function, while the second utilizes hybrid effective field theory (HEFT) to reproduce these quantities within the simulation. Our analysis involves comparing both methods to the true MillenniumTNG optical depth and Compton-y fields using several metrics, including the cross-correlation coefficient, power spectrum, and power spectrum error. Additionally, we assess how well the reconstructed fields correlate with dark matter haloes across various mass thresholds. The results indicate that the transfer function method yields more accurate reconstructions for fields with initially high correlations (r ≈ 1), such as between the optical depth and dark matter fields. Conversely, the HEFT-based approach proves more effective in enhancing correlations for fields with weaker initial correlations (r ∼ 0.5), such as between the Compton-y and dark matter fields. Lastly, we discuss extensions of our methods to improve the reconstruction performance at the field level.

Liu, R. Henry [University of California, Berkeley,↗

A Surrogate-based Adaptive Sampling Approach for History Matching and Uncertainty Quantification

A critical procedure in reservoir simulations is history matching (or data assimilation in a broader sense), which calibrates model parameters such that the simulation results are consistent with field measurements, and hence improves the credibility of the predictions given by the simulations. Often there exist non-unique combinations of parameter values that all yield the simulation results matching the measurements. For such ill-posed history matching problems, Bayesian theorem provides a theoretical foundation to represent different solutions and to quantify the uncertainty with the posterior PDF. Lacking an analytical solution in most situations, the posterior PDF may be characterized with a sample of realizations, each representing a possible scenario. A novel sampling algorithm is presented here for the Bayesian solutions to history matching problems. We aim to deal with two commonly encountered issues: 1) as a result of the nonlinear input-output relationship in a reservoir model, the posterior distribution could be in a complex form, such as multimodal, which violates the Gaussian assumption required by most of the commonly used data assimilation approaches; 2) a typical sampling method requires intensive model evaluations and hence may cause unaffordable computational cost. In the developed algorithm, we use a Gaussian mixture model as the proposal distribution in the sampling process, which is simple but also flexible to approximate non-Gaussian distributions and is particularly efficient when the posterior is multimodal. Also, a Gaussian process is utilized as a surrogate model to speed up the sampling process. Furthermore, an iterative scheme of adaptive surrogate refinement and re-sampling ensures sampling accuracy while keeping the computational cost at a minimum level. The developed approach is demonstrated with an illustrative example and shows its capability in handling the above-mentioned issues. Multimodal posterior of the history matching problem is captured and are used to give a reliable production prediction with uncertainty quantification. The new algorithm reveals a great improvement in terms of computational efficiency comparing previously studied approaches for the sample problem.

Li, Weixuan↗

Performance of Solar-Induced Chlorophyll Fluorescence in Estimating Water-Use Efficiency in a Temperate Forest

Water-use efficiency (WUE) is a critical variable describing the interrelationship between carbon uptake and water loss in land ecosystems. Different WUE formulations (WUEs) including intrinsic water use efficiency (WUEi), inherent water use efficiency (IWUE), and underlying water use efficiency (uWUE) have been proposed. Based on continuous measurements of carbon and water fluxes and solar-induced chlorophyll fluorescence (SIF) at a temperate forest, we analyze the correlations between SIF emission and the different WUEs at the canopy level by using linear regression (LR) and Gaussian processes regression (GPR) models. Overall, we find that SIF emission has a good potential to estimate IWUE and uWUE, especially when a combination of different SIF bands and a GPR model is used. At an hourly time step, canopy-level SIF emission can explain as high as 65% and 61% of the variances in IWUE and uWUE. Specifically, we find that (1) a daily time step by averaging hourly values during daytime can enhance the SIF-IWUE correlations, (2) the SIF-IWUE correlations decrease when photosynthetically active radiation and air temperature exceed their optimal biological thresholds, (3) a low Leaf Area Index (LAI) has a negative effect on the SIF-IWUE correlations due to large evaporation fluxes, (4) a high LAI in summer also reduces the SIF-IWUE correlations most likely due to increasing scattering and (re)absorption of the SIF signal, and (5) the observation time during the day has a strong impact on the SIF-IWUE correlations and SIF measurements in the early morning have the lowest power to estimate IWUE due to the large evaporation of dew. This study provides a new way to evaluate the stomatal regulation of plant-gas exchange without complex parameterizations.

54 ENVIRONMENTAL SCIENCES↗

Towards fast and accurate predictions of radio frequency power deposition and current profile via data-driven modelling: applications to lower hybrid current drive

Three machine learning techniques (multilayer perceptron, random forest and Gaussian process) provide fast surrogate models for lower hybrid current drive (LHCD) simulations. A single GENRAY/CQL3D simulation without radial diffusion of fast electrons requires several minutes of wall-clock time to complete, which is acceptable for many purposes, but too slow for integrated modelling and real-time control applications. The machine learning models use a database of more than 16 000 GENRAY/CQL3D simulations for training, validation and testing. Latin hypercube sampling methods ensure that the database covers the range of nine input parameters ( $n_{e0}$ , $T_{e0}$ , $I_p$ , $B_t$ , $R_0$ , $n_{\|}$ , $Z_{{\rm eff}}$ , $V_{{\rm loop}}$ and $P_{{\rm LHCD}}$ ) with sufficient density in all regions of parameter space. The surrogate models reduce the inference time from minutes to $\sim$ ms with high accuracy across the input parameter space.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Combining multitask and transfer learning with deep Gaussian processes for autotuning-based performance engineering

We combine deep Gaussian processes (DGPs) with multitask and transfer learning for the performance modeling and optimization of HPC applications. Deep Gaussian processes merge the uncertainty quantification advantage of Gaussian processes (GPs) with the predictive power of deep learning. Multitask and transfer learning allow for improved learning efficiency when several similar tasks are to be learned simultaneously and when previous learned models are sought to help in the learning of new tasks, respectively. A comparison with state-of-the-art autotuners shows the advantage of our approach on two application problems. In this article, we combine DGPs with multitask and transfer learning to allow for both an improved tuning of an application parameters on problems of interest but also the prediction of parameters on any potential problem the application might encounter.

97 MATHEMATICS AND COMPUTING↗