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 631 records · Page 35

PARAMETER ESTIMATION FROM TIME-SERIES DATA WITH CORRELATED ERRORS: A WAVELET-BASED METHOD AND ITS APPLICATION TO TRANSIT LIGHT CURVES

We consider the problem of fitting a parametric model to time-series data that are afflicted by correlated noise. The noise is represented by a sum of two stationary Gaussian processes: one that is uncorrelated in time, and another that has a power spectral density varying as 1/f{sup g}amma. We present an accurate and fast [O(N)] algorithm for parameter estimation based on computing the likelihood in a wavelet basis. The method is illustrated and tested using simulated time-series photometry of exoplanetary transits, with particular attention to estimating the mid-transit time. We compare our method to two other methods that have been used in the literature, the time-averaging method and the residual-permutation method. For noise processes that obey our assumptions, the algorithm presented here gives more accurate results for mid-transit times and truer estimates of their uncertainties.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Uncertainty Quantification for Data-Driven Machine Learning Models in Nuclear Engineering Applications: Where We Are and What Do We Need?

Machine learning (ML) has been leveraged to tackle a diverse range of tasks in almost all branches of nuclear engineering. Many of the successes in ML applications can be attributed to the recent performance breakthroughs in deep learning, the growing availability of computational power, data, and easy-to-use ML libraries. However, these empirical successes have often outpaced our formal understanding of the ML algorithms. An important but under-rated area is uncertainty quantification (UQ) of ML. ML-based models are subject to approximation uncertainty when they are used to make predictions, due to sources including but not limited to, data noise, data coverage, extrapolation, imperfect model architecture and the stochastic training process. The goal of this paper is to clearly explain and illustrate the importance of UQ of ML. We will elucidate the differences in the basic concepts of UQ of physics-based models and data-driven ML models. Various sources of uncertainties in physical modeling and data-driven modeling will be discussed, demonstrated, and compared. We will also present and demonstrate a few techniques to quantify the ML prediction uncertainties, including Monte Carlo dropout, deep ensemble, Bayesian neural networks, Gaussian Processes and conformal prediction. Lastly, we will discuss the need for building a verification, validation and UQ framework to establish ML credibility.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Reinforcement Learning via Gaussian Processes with Neural Network Dual Kernels

While deep neural networks (DNNs) and Gaussian Processes (GPs) are both popularly utilized to solve problems in reinforcement learning, both approaches feature undesirable drawbacks for challenging problems. DNNs learn complex non-linear embeddings, but do not naturally quantify uncertainty and are often data-inefficient to train. GPs infer posterior distributions over functions, but popular kernels exhibit limited expressivity on complex and high-dimensional data. Fortunately, recently discovered conjugate and neural tangent kernel functions encode the behavior of overparameterized neural networks in the kernel domain. We demonstrate that these kernels can be efficiently applied to regression and reinforcement learning problems by analyzing a baseline case study.We apply GPs with neural network dual kernels to solve reinforcement learning tasks for the first time. We demonstrate, using the well understood mountain-car problem, that GPs empowered with dual kernels perform at least as well as those using the conventional radial basis function kernel. Finally, we conjecture that by inheriting the probabilistic rigor of GPs and the powerful embedding properties of DNNs, GPs using NN dual kernels will empower future reinforcement learning models on difficult domains.

97 MATHEMATICS AND COMPUTING↗

A Simple Model of the Variability of Soil Depths

Soil depth tends to vary from a few centimeters to several meters, depending on many natural and environmental factors. Here, we hypothesize that the cumulative effect of these factors on soil depth, which is chiefly dependent on the process of biogeochemical weathering, is particularly affected by soil porewater (i.e., solute) transport and infiltration from the land surface. Taking into account evidence for a non-Gaussian distribution of rock weathering rates, we propose a simple mathematical model to describe the relationship between soil depth and infiltration flux. The model was tested using several areas in mostly semi-arid climate zones. The application of this model demonstrates the use of fundamental principles of physics to quantify the coupled effects of the five principal soil-forming factors of Dokuchaev.

54 ENVIRONMENTAL SCIENCES↗

Short-Term Load Forecasting Considering EV Charging Loads with Prediction Interval Evaluation

Short-term load forecasting plays a critical role in power system planning and operation. Along with the electrification of various loads, electricity demands are becoming increasingly hard to predict. Notably, the recent rise in electric vehicles (EVs) has further contributed to this unpredictability. To address this issue, this paper proposes a probabilistic load forecasting strategy utilizing Gaussian process regression, structured in a day-ahead manner. While many works focus on deterministic prediction, probabilistic forecasting offers additional insights into variability and uncertainty, enabling more flexible and reliable operation for power systems. To enhance the accuracy of the load forecasting model, the inputs include features related to EV charging habits as well as commonly used weather information. The load forecasting results are evaluated using various metrics, including conventional ones that assess the accuracy of point forecasts, as well as additional metrics that test the reliability of prediction intervals. The proposed load forecasting method is finally tested on real residential power consumption data and EV charging data sampled from real-world sources. The results prove that the new features can greatly improve the performance of the load forecasting method.

electrical vehicle↗

High-speed spiral imaging technique for an atomic force microscope using a linear quadratic Gaussian controller

This paper demonstrates a high-speed spiral imaging technique for an atomic force microscope (AFM). As an alternative to traditional raster scanning, an approach of gradient pulsing using a spiral line is implemented and spirals are generated by applying single-frequency cosine and sine waves of slowly varying amplitudes to the X and Y-axes of the AFM’s piezoelectric tube scanner (PTS). Due to these single-frequency sinusoidal input signals, the scanning process can be faster than that of conventional raster scanning. A linear quadratic Gaussian controller is designed to track the reference sinusoid and a vibration compensator is combined to damp the resonant mode of the PTS. An internal model of the reference sinusoidal signal is included in the plant model and an integrator for the system error is introduced in the proposed control scheme. As a result, the phase error between the input and output sinusoids from the X and Y-PTSs is reduced. The spirals produced have particularly narrow-band frequency measures which change slowly over time, thereby making it possible for the scanner to achieve improved tracking and continuous high-speed scanning rather than being restricted to the back and forth motion of raster scanning. As part of the post-processing of the experimental data, a fifth-order Butterworth filter is used to filter noises in the signals emanating from the position sensors and a Gaussian image filter is used to filter the images. A comparison of images scanned using the proposed controller (spiral) and the AFM PI controller (raster) shows improvement in the scanning rate using the proposed method.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

SUGAR: An improved empirical model of Type Ia supernovae based on spectral features

Type Ia supernovae (SNe Ia) are widely used to measure the expansion of the Universe. Improving distance measurements of SNe Ia is one technique to better constrain the acceleration of expansion and determine its physical nature. This document develops a new SNe Ia spectral energy distribution (SED) model, called the SUpernova Generator And Reconstructor (SUGAR), which improves the spectral description of SNe Ia, and consequently could improve the distance measurements. This model was constructed from SNe Ia spectral properties and spectrophotometric data from the Nearby Supernova Factory collaboration. In a first step, a principal component analysis-like method was used on spectral features measured at maximum light, which allowed us to extract the intrinsic properties of SNe Ia. Next, the intrinsic properties were used to extract the average extinction curve. Third, an interpolation using Gaussian processes facilitated using data taken at different epochs during the lifetime of an SN Ia and then projecting the data on a fixed time grid. Finally, the three steps were combined to build the SED model as a function of time and wavelength. This is the SUGAR model. Results. The main advancement in SUGAR is the addition of two additional parameters to characterize SNe Ia variability. The first is tied to the properties of SNe Ia ejecta velocity and the second correlates with their calcium lines. The addition of these parameters, as well as the high quality of the Nearby Supernova Factory data, makes SUGAR an accurate and efficient model for describing the spectra of normal SNe Ia as they brighten and fade. The performance of this model makes it an excellent SED model for experiments like the Zwicky Transient Facility, the Large Synoptic Survey Telescope, or the Wide Field Infrared Survey Telescope.

79 ASTRONOMY AND ASTROPHYSICS↗

Theoretical Analysis of the EPR Spectrum Saturation Effect with Account for Spectral Diffusion in a System with a Gaussian Distribution of Spin Resonance Frequencies

We have analyzed the effect of saturation of the EPR spectrum for the model situation when a set of resonance frequencies of spin packets is described by a Gaussian distribution and spectral diffusion, which is a random process without correlation, exists. We have obtained the form of the spectrum for arbitrary values of magnetic–resonant parameters: relaxation times for the longitudinal and transverse components of the magnetization vector of the system, the spectral diffusion rate, the dispersion of the Gaussian distribution of resonance frequency, and the magnetic induction of the microwave field. We have derived an analytic expression for the saturation curve of the EPR spectrum. Detailed analysis of the dependence of the spectrum saturation curve on the spin system parameters is performed for the optimal case when the microwave field frequency coincides with the average frequency of the Gaussian distribution for resonance frequencies. The value of the microwave field induction at which the peak of the saturation curve is attained has been determined for arbitrary values of the paramagnetic relaxation time, spectral diffusion rate, and dispersion of the Gaussian spin frequency distribution. We have formulated the algorithm for determining the spin–lattice relaxation time using the experimentally determined microwave field induction for which the saturation curve passes through the peak. It should be noted that numerical calculations have been performed for parameters typical of EPR spectroscopy. The general analytic expression derived for the shape of the spectrum in the saturation conditions and qualitative conclusions concerning the effect of spectral diffusion on the saturation of stationary spectra obtained in this study can also be used in other fields of spectroscopy.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Ion Trapping Studies and Mitigation Strategies for the EIC ERL-Based Strong Hadron Cooler

An Energy Recovery Linac based strong hadron cooler was previously considered for the Electron-Ion Collider. The required electron beam parameters for variable-energy strong hadron cooling place significant constraints on ion trapping and collective effects. This paper presents initial studies of these constraints through a combination of analytical modelling and numerical simulations of ion production, trapping behaviour, and mitigation strategies. A multi-bunch tracking framework based on ELEGANT with the ionEffects module is used to simulate machine operation over millisecond time scales, corresponding to more than 3 × 10^5 electron bunches. The simulations include modelling of ionisation processes together with transverse electron–ion dynamics, allowing the evolution and accumulation of ions to be investigated. Analytical expressions based on Gaussian beam distributions are used to estimate ion trapping conditions and benchmark the simulation results. A bi-periodic bunch spacing scheme is also investigated as a possible mitigation method by detuning the ion oscillation frequency. These studies provide an initial assessment of ion trapping in the strong hadron cooler and demonstrate possible approaches for reducing beam–ion effects.

Bi, R. [Lancaster University, Cockcroft Institute]↗

Enter Gaussian Mixture Modeling Extensions for Improved False Discovery Rate Estimation in GC-MS Metabolomics

Identifying small molecules (e.g., metabolites) is key towards driving scientific advancement in metabolomics, and gas chromatography–mass spectrometry (GC-MS) is an analytic method that may be applied to facilitate this process. The typical GC-MS identification workflow involves quantifying the similarity of an observed sample spectrum and other features (e.g. retention index) to that of several references, noting the compound of the best-matching reference spectrum as the identified metabolite. While a deluge of similarity metrics exists, none characterize the error rate of generated identifications, thereby presenting an unknown risk of false identification or discovery. To quantify this unknown risk, we propose a model-based framework for estimating the false discovery rate (FDR) among a set of identifications. Extending the traditional mixture modeling framework, our method incorporates both similarity score and experimental information in estimating the FDR. We apply these models to identification lists derived from across 548 samples of varying complexity and sample type (e.g., fungal species, standard mixtures, etc.), comparing their performance to that of the traditional Gaussian mixture model (GMM). Through simulation, we additionally assess the impact of reference library size on the accuracy of FDR estimates. In comparing the best performing model extensions to the GMM, our results indicate relative decreases in median absolute estimation error (MAE) ranging from 12% to 70%, based on comparisons of the median MAEs across all hit-lists. Results indicate that these relative performance improvements generally hold despite library size, however FDR estimation error typically worsens as the set of reference compounds diminishes.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Fast and accurate reduced-order modeling of a MOOSE-based additive manufacturing model with operator learning

One predominant challenge in additive manufacturing (AM) is to achieve specific material properties by manipulating manufacturing process parameters during the runtime. Such manipulation tends to increase the computational load imposed on existing simulation tools employed in AM. The goal of the present work is to construct a fast and accurate reduced-order model (ROM) for an AM model developed within the Multiphysics Object-Oriented Simulation Environment (MOOSE) framework, ultimately reducing the time/cost of AM control and optimization processes. Our adoption of the operator learning (OL) approach enabled us to learn a family of differential equations produced by altering process variables in the laser’s Gaussian point heat source. More specifically, we used the Fourier neural operator (FNO) and deep operator network (DeepONet) to develop ROMs for time-dependent responses. Furthermore, we benchmarked the performance of these OL methods against a conventional deep neural network (DNN)-based ROM. Ultimately, we found that OL methods offer comparable performance and, in terms of accuracy and generalizability, even outperform DNN at predicting scalar model responses. The DNN-based ROM afforded the fastest training time. Furthermore, all the ROMs were faster than the original MOOSE model yet still provided accurate predictions. FNO had a smaller mean prediction error than DeepONet, with a larger variance for time-dependent responses. Unlike DNN, both FNO and DeepONet were able to simulate time series data without the need for dimensionality reduction techniques. Finally, the present work can help facilitate the AM optimization process by enabling faster execution of simulation tools while still preserving evaluation accuracy.

36 MATERIALS SCIENCE↗

Jet fragmentation transverse momentum distributions in pp and p-Pb collisions at $ \sqrt{s}$, $\sqrt{s_{\mathrm{NN}}}$ = 5.02 TeV

Jet fragmentation transverse momentum ( j T ) distributions are measured in proton-proton (pp) and proton-lead (p-Pb) collisions at s NN = 5 . 02 TeV with the ALICE experiment at the LHC. Jets are reconstructed with the ALICE tracking detectors and electromagnetic calorimeter using the anti- k T algorithm with resolution parameter R = 0 . 4 in the pseudorapidity range |η| < 0 . 25. The j T values are calculated for charged particles inside a fixed cone with a radius R = 0 . 4 around the reconstructed jet axis. The measured j T distributions are compared with a variety of parton-shower models. Herwig and P ythia 8 based models describe the data well for the higher j T region, while they underestimate the lower j T region. The j T distributions are further characterised by fitting them with a function composed of an inverse gamma function for higher j T values (called the “wide component”), related to the perturbative component of the fragmentation process, and with a Gaussian for lower j T values (called the “narrow component”), predominantly connected to the hadronisation process. The width of the Gaussian has only a weak dependence on jet transverse momentum, while that of the inverse gamma function increases with increasing jet transverse momentum. For the narrow component, the measured trends are successfully described by all models except for Herwig. For the wide component, Herwig and PYTHIA 8 based models slightly underestimate the data for the higher jet transverse momentum region. These measurements set constraints on models of jet fragmentation and hadronisation.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Statistical characteristics of cloud variability. Part 2: Implication for parameterizations of microphysical and radiative transfer processes in climate models

The effects of subgrid cloud variability on grid-average microphysical rates and radiative fluxes are examined by use of long-term retrieval products at the Tropical West Pacific, Southern Great Plains, and North Slope of Alaska sites of the Department of Energy's Atmospheric Radiation Measurement program. Four commonly used distribution functions, the truncated Gaussian, Gamma, lognormal, and Weibull distributions, are constrained to have the same mean and standard deviation as observed cloud liquid water content. The probability density functions are then used to upscale relevant physical processes to obtain grid-average process rates. It is found that the truncated Gaussian representation results in up to 30% mean bias in autoconversion rate, whereas the mean bias for the lognormal representation is about 10%. The Gamma and Weibull distribution function performs the best for the grid-average autoconversion rate with the mean relative bias less than 5%. For radiative fluxes, the lognormal and truncated Gaussian representations perform better than the Gamma and Weibull representations. The results show that the optimal choice of subgrid cloud distribution function depends on the nonlinearity of the process of interest, and thus, there is no single distribution function that works best for all parameterizations. Examination of the scale (window size) dependence of the mean bias indicates that the bias in grid-average process rates monotonically increases with increasing window sizes, suggesting the increasing importance of subgrid variability with increasing grid sizes.

54 ENVIRONMENTAL SCIENCES↗

Simultaneous Optical Transmission Spectroscopy of a Terrestrial, Habitable-zone Exoplanet with Two Ground-based Multiobject Spectrographs

Investigating the atmospheres of rocky exoplanets is key to performing comparative planetology between these worlds and the terrestrial planets that reside in the inner solar system. Terrestrial exoplanet atmospheres exhibit weak signals, and attempting to detect them pushes at the boundaries of what is possible for current instrumentation. We focus on the habitable-zone terrestrial exoplanet LHS 1140b. Given its 25-day orbital period and 2 hr transit duration, capturing transits of LHS 1140b is challenging. We observed two transits of this object, approximately 1 yr apart, which yielded four data sets thanks to our simultaneous use of the IMACS and LDSS3C multiobject spectrographs mounted on the twin Magellan telescopes at Las Campanas Observatory. We present a jointly fit white light curve, as well as jointly fit 20 nm wavelength-binned light curves from which we construct a transmission spectrum. Binning the joint white light-curve residuals to 3-minute time bins gives an rms of 145 ppm; binning down to 10-minute time bins gives an rms of 77 ppm. Our median uncertainty in R{sub p}{sup 2}/R{sub s}{sup 2} in the 20 nm wavelength bins is 260 ppm, and we achieve an average precision of 1.3× the photon noise when fitting the wavelength-binned light curves with a Gaussian process regression. Our precision on R{sub p}{sup 2}/R{sub s}{sup 2} is a factor of four larger than the feature amplitudes of a clear, hydrogen-dominated atmosphere, meaning that we are not able to test realistic models of LHS 1140b’s atmosphere. The techniques and caveats presented here are applicable to the growing sample of terrestrial worlds in the Transiting Exoplanet Survey Satellite era, as well as to the upcoming generation of ground-based giant segmented mirror telescopes.

79 ASTRONOMY AND ASTROPHYSICS↗

Recipes for when physics fails: recovering robust learning of physics informed neural networks

Abstract Physics-informed neural networks (PINNs) have been shown to be effective in solving partial differential equations by capturing the physics induced constraints as a part of the training loss function. This paper shows that a PINN can be sensitive to errors in training data and overfit itself in dynamically propagating these errors over the domain of the solution of the PDE. It also shows how physical regularizations based on continuity criteria and conservation laws fail to address this issue and rather introduce problems of their own causing the deep network to converge to a physics-obeying local minimum instead of the global minimum. We introduce Gaussian process (GP) based smoothing that recovers the performance of a PINN and promises a robust architecture against noise/errors in measurements. Additionally, we illustrate an inexpensive method of quantifying the evolution of uncertainty based on the variance estimation of GPs on boundary data. Robust PINN performance is also shown to be achievable by choice of sparse sets of inducing points based on sparsely induced GPs. We demonstrate the performance of our proposed methods and compare the results from existing benchmark models in literature for time-dependent Schrödinger and Burgers’ equations.

97 MATHEMATICS AND COMPUTING↗

Surface enrichment dictates block copolymer orientation

Orientation of block copolymer (BCP) morphology in thin films is critical to applications as nanostructured coatings. Despite being well-studied, the ability to control BCP orientation across all possible block constituents remains challenging. Here, in this study, we deploy coarse-grained molecular dynamics simulations to study diblock copolymer ordering in thin films, focusing on chain makeup, substrate surface energy, and surface tension disparity between the two constituent blocks. We explore the multi-dimensional parameter space of ordering using a machine-learning approach, where an autonomous loop using a Gaussian process (GP) control algorithm iteratively selects high-value simulations to compute. The GP kernel was engineered to capture known symmetries. The trained GP model serves as both a complete map of system response, and a robust means of extracting material knowledge. We demonstrate that the vertical orientation of BCP phases depends on several counter-balancing energetic contributions, including entropic and enthalpic material enrichment at interfaces, distortion of morphological objects through the film depth, and of course interfacial energies. BCP lamellae are found more resistant to these effects, and thus more robustly form vertical orientations across a broad range of conditions; while BCP cylinders are found to be highly sensitive to surface tension disparity.

77 NANOSCIENCE AND NANOTECHNOLOGY↗

Generalized uncertainty relations and entanglement dynamics in quantum Brownian motion models

We study entanglement dynamics in quantum Brownian motion (QBM) models. Our main tool is the Wigner function propagator. Time evolution in the Wigner picture is physically intuitive and it leads to a simple derivation of a master equation for any number of system harmonic oscillators and spectral density of the environment. It also provides generalized uncertainty relations, valid for any initial state, that allow a characterization of the environment in terms of the modifications it causes to the system's dynamics. In particular, the uncertainty relations are very informative about the entanglement dynamics of Gaussian states, and to a lesser extent for other families of states. For concreteness, we apply these techniques to a bipartite QBM model, describing the processes of entanglement creation, disentanglement, and decoherence at all temperatures and time scales.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Efficient near-field ptychography reconstruction using the Hessian operator

X-ray ptychography is a powerful and robust coherent imaging method providing access to the complex object and probe (illumination). Ptychography reconstruction is typically performed using first-order methods due to their computational efficiency. Higher-order methods, while potentially more accurate, are often prohibitively expensive in terms of computation. In this study, we present a mathematical framework for reconstruction using second-order information derived from an efficient computation of the bilinear Hessian and Hessian operator. The formulation is provided for Gaussian-based models, enabling the simultaneous reconstruction of the object, probe, and object positions. Synthetic data tests, along with experimental near-field ptychography data processing, demonstrate a ten-fold reduction in computation time compared to first-order methods. The derived formulas for computing the Hessians, along with the strategies for incorporating them into optimization schemes, are well-structured and easily adaptable to various ptychography problem formulations.

Carlsson, Marcus [Lund Univ. (Sweden)] (ORCID:0000↗