SEARCH · Search NASA
Results for “inference methods”
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
SU-F-J-34: Automatic Target-Based Patient Positioning Framework for Image-Guided Radiotherapy in Prostate Cancer Treatment
Purpose: Current image-guided radiotherapy (IGRT) procedure is bonebased patient positioning, followed by subjective manual correction using cone beam computed tomography (CBCT). This procedure might cause the misalignment of the patient positioning. Automatic target-based patient positioning systems achieve the better reproducibility of patient setup. Our aim of this study was to develop an automatic target-based patient positioning framework for IGRT with CBCT images in prostate cancer treatment. Methods: Seventy-three CBCT images of 10 patients and 24 planning CT images with digital imaging and communications in medicine for radiotherapy (DICOM-RT) structures were used for this study. Our proposed framework started from the generation of probabilistic atlases of bone and prostate from 24 planning CT images and prostate contours, which were made in the treatment planning. Next, the gray-scale histograms of CBCT values within CTV regions in the planning CT images were obtained as the occurrence probability of the CBCT values. Then, CBCT images were registered to the atlases using a rigid registration with mutual information. Finally, prostate regions were estimated by applying the Bayesian inference to CBCT images with the probabilistic atlases and CBCT value occurrence probability. The proposed framework was evaluated by calculating the Euclidean distance of errors between two centroids of prostate regions determined by our method and ground truths of manual delineations by a radiation oncologist and a medical physicist on CBCT images for 10 patients. Results: The average Euclidean distance between the centroids of extracted prostate regions determined by our proposed method and ground truths was 4.4 mm. The average errors for each direction were 1.8 mm in anteroposterior direction, 0.6 mm in lateral direction and 2.1 mm in craniocaudal direction. Conclusion: Our proposed framework based on probabilistic atlases and Bayesian inference might be feasible to automatically determine prostate regions on CBCT images.
Discovery of Probabilistic Dirichlet-to-Neumann Maps on Graphs
Dirichlet-to-Neumann maps enable the coupling of multiphysics simulations across computational subdomains by ensuring continuity of state variables and fluxes at artificial interfaces. We present a novel method for learning Dirichlet-to-Neumann maps on graphs using Gaussian processes, specifically for problems where the data obey a conservation law arising from an underlying partial differential equation. Our approach combines discrete exterior calculus and nonlinear optimal recovery to infer relationships between vertex and edge values. This framework yields data-driven predictions with uncertainty quantification across the entire graph, even when observations are limited to a subset of vertices and edges. By minimizing the reproducing kernel Hilbert space norm while penalizing kernel complexity through maximum likelihood estimation, our method ensures that the resulting surrogate strictly enforces conservation laws without overfitting. We demonstrate our method on two representative applications: subsurface flow in fracture networks and arterial blood flow. Finally, the results demonstrate that the method maintains high accuracy and well-calibrated uncertainty estimates even under severe data scarcity, highlighting its potential for scientific applications where limited data and reliable uncertainty quantification are critical.
Report of the Workshop on Advanced Fusion with Machine Learning, April 30-May 2, 2019
The pursuit of fusion energy has led to extensive activities in both experimental and theoretical science. The central goal of all these activities has been to develop the knowledge necessary for the design of successful fusion power plants. Even today, following decades of research in many key areas including plasma physics and material science, much remains to be learned in order to achieve optimization of the tokamak or other paths to fusion energy. Data science methods from the fields of machine learning and artificial intelligence (ML/AI) offer opportunities for enabling or accelerating progress toward the realization of fusion energy by maximizing the amount and usefulness of information extracted from experimental and simulation output data. From April 30 to May 2 of 2019, a workshop on “Advancing Fusion with Machine Learning Research Needs” was held in Gaithersburg, MD, with joint support from the Fusion Energy Science (FES) and Advanced Scientific Computing Research (ASCR) Programs. The workshop brought together ~ 60 experts in fields spanning fusion science, data science, statistical inference and mathematics, machine learning, and artificial intelligence, along with DOE program managers and technical experts, with the goal of identifying Priority Research Opportunities (PROs) for the application of ML/AI methods to accelerating progress on solving fusion problems. During the workshop, seven key PROs were identified. The description of each PRO includes relevant fusion problems, potentially useful ML/AI approaches, known gaps currently preventing the use of such approaches, and research guidelines to maximize the effective application of ML/AI methods to the PRO. The report containing the full description of the workshop outcomes is available at https://science.osti.gov/-/media/fes/pdf/workshopreports/FES_ASCR_Machine_Learning_Report.pdf
A flexible class of priors for orthonormal matrices with basis function-specific structure
Statistical modeling of high-dimensional matrix-valued data motivates the use of a low-rank representation that simultaneously summarizes key characteristics of the data and enables dimension reduction. Low-rank representations commonly factor the original data into the product of orthonormal basis functions and weights, where each basis function represents an independent feature of the data. However, the basis functions in these factorizations are typically computed using algorithmic methods that cannot quantify uncertainty or account for basis function correlation structure a priori. While there exist Bayesian methods that allow for a common correlation structure across basis functions, empirical examples motivate the need for basis function-specific dependence structure. We propose a prior distribution for orthonormal matrices that can explicitly model basis function-specific structure. The prior is used within a general probabilistic model for singular value decomposition to conduct posterior inference on the basis functions while accounting for measurement error and fixed effects. We discuss how the prior specification can be used for various scenarios and demonstrate favorable model properties through synthetic data examples. Finally, we apply our method to two-meter air temperature data from the Pacific Northwest, enhancing our understanding of the Earth system’s internal variability.
Implementation of a practical Markov chain Monte Carlo sampling algorithm in PyBioNetFit
Abstract Summary Bayesian inference in biological modeling commonly relies on Markov chain Monte Carlo (MCMC) sampling of a multidimensional and non-Gaussian posterior distribution that is not analytically tractable. Here, we present the implementation of a practical MCMC method in the open-source software package PyBioNetFit (PyBNF), which is designed to support parameterization of mathematical models for biological systems. The new MCMC method, am, incorporates an adaptive move proposal distribution. For warm starts, sampling can be initiated at a specified location in parameter space and with a multivariate Gaussian proposal distribution defined initially by a specified covariance matrix. Multiple chains can be generated in parallel using a computer cluster. We demonstrate that am can be used to successfully solve real-world Bayesian inference problems, including forecasting of new Coronavirus Disease 2019 case detection with Bayesian quantification of forecast uncertainty. Availability and implementation PyBNF version 1.1.9, the first stable release with am, is available at PyPI and can be installed using the pip package-management system on platforms that have a working installation of Python 3. PyBNF relies on libRoadRunner and BioNetGen for simulations (e.g. numerical integration of ordinary differential equations defined in SBML or BNGL files) and Dask.Distributed for task scheduling on Linux computer clusters. The Python source code can be freely downloaded/cloned from GitHub and used and modified under terms of the BSD-3 license (https://github.com/lanl/pybnf). Online documentation covering installation/usage is available (https://pybnf.readthedocs.io/en/latest/). A tutorial video is available on YouTube (https://www.youtube.com/watch?v=2aRqpqFOiS4&t=63s). Supplementary information Supplementary data are available at Bioinformatics online.
Plaster: An Integration, Benchmark and Development Framework for Heterogeneous Metadata Normalization Methods
The recent advances in the automation of metadata normalization and the invention of a unified schema — Brick — alleviate the metadata normalization challenge for deploying portable applications across buildings. Yet, the lack of compatibility between existing metadata normalization methods precludes the possibility of comparing and combining them. While generic machine learning (ML) frameworks, such as MLJAR and OpenML, provide versatile interfaces for standard ML problems, they cannot easily accommodate the metadata normalization tasks for buildings due to the heterogeneity in the inference scope, type of data required as input, evaluation metric, and the building-specific human-in-the-loop learning procedure. We propose Plaster, an open and modular framework that incorporates existing advances in building metadata normalization. It provides unified programming interfaces for various types of learning methods for metadata normalization and defines standardized data models for building metadata and timeseries data. Thus, it enables the integration of different methods via a workflow, benchmarking of different methods via unified interfaces, and rapid prototyping of new algorithms. With Plaster, we 1) show three examples of the workflow integration, delivering better performance than individual algorithms, 2) benchmark/analyze five algorithms over five common buildings, and 3) exemplify the process of developing a new algorithm involving time series features. We believe Plaster will facilitate the development of new algorithms and expedite the adoption of standard metadata schema such as Brick, in order to enable seamless smart building applications in the future.
Atomistic Method Applied to Computational Modeling of Surface Alloys
The formation of surface alloys is a growing research field that, in terms of the surface structure of multicomponent systems, defines the frontier both for experimental and theoretical techniques. Because of the impact that the formation of surface alloys has on surface properties, researchers need reliable methods to predict new surface alloys and to help interpret unknown structures. The structure of surface alloys and when, and even if, they form are largely unpredictable from the known properties of the participating elements. No unified theory or model to date can infer surface alloy structures from the constituents properties or their bulk alloy characteristics. In spite of these severe limitations, a growing catalogue of such systems has been developed during the last decade, and only recently are global theories being advanced to fully understand the phenomenon. None of the methods used in other areas of surface science can properly model even the already known cases. Aware of these limitations, the Computational Materials Group at the NASA Glenn Research Center at Lewis Field has developed a useful, computationally economical, and physically sound methodology to enable the systematic study of surface alloy formation in metals. This tool has been tested successfully on several known systems for which hard experimental evidence exists and has been used to predict ternary surface alloy formation (results to be published: Garces, J.E.; Bozzolo, G.; and Mosca, H.: Atomistic Modeling of Pd/Cu(100) Surface Alloy Formation. Surf. Sci., 2000 (in press); Mosca, H.; Garces J.E.; and Bozzolo, G.: Surface Ternary Alloys of (Cu,Au)/Ni(110). (Accepted for publication in Surf. Sci., 2000.); and Garces, J.E.; Bozzolo, G.; Mosca, H.; and Abel, P.: A New Approach for Atomistic Modeling of Pd/Cu(110) Surface Alloy Formation. (Submitted to Appl. Surf. Sci.)). Ternary alloy formation is a field yet to be fully explored experimentally. The computational tool, which is based on the BFS (Bozzolo, Ferrante, and Smith) method for the calculation of the energetics, consists of a small number of simple PCbased computer codes that deal with the different aspects of surface alloy formation. Two analysis modes are available within this package. The first mode provides an atom-by-atom description of real and virtual stages 1. during the process of surface alloying, based on the construction of catalogues of configurations where each configuration describes one possible atomic distribution. BFS analysis of this catalogue provides information on accessible states, possible ordering patterns, and details of island formation or film growth. More importantly, it provides insight into the evolution of the system. Software developed by the Computational Materials Group allows for the study of an arbitrary number of elements forming surface alloys, including an arbitrary number of surface atomic layers. The second mode involves large-scale temperature-dependent computer 2. simulations that use the BFS method for the energetics and provide information on the dynamic processes during surface alloying. These simulations require the implementation of Monte-Carlo-based codes with high efficiency within current workstation environments. This methodology capitalizes on the advantages of the BFS method: there are no restrictions on the number or type of elements or on the type of crystallographic structure considered. This removes any restrictions in the definition of the configuration catalogues used in the analytical calculations, thus allowing for the study of arbitrary ordering patterns, ultimately leading to the actual surface alloy structure. Moreover, the Monte Carlo numerical technique used for the large-scale simulations allows for a detailed visualization of the simulated process, the main advantage of this type of analysis being the ability to understand the underlying features that drive these processes. Because of the simplicity of the BFS method for e energetics used in these calculations, a detailed atom-by-atom analysis can be performed at any point in the simulation, providing necessary insight on the details of the process. The main objective of this research program is to develop a tool to guide experimenters in understanding and interpreting often unexpected results in alloy formation experiments. By reducing the computational effort without losing physical accuracy, we expect that powerful simulation tools will be developed in the immediate future, which will allow material scientists to easily visualize and analyze processes at a level not achievable experimentally.
Developing a laser shockwave model for characterizing diffusion bonded interfaces
The US National Nuclear Security Agency has a Global Threat Reduction Initiative (GTRI) with the goal of reducing the worldwide use of high-enriched uranium (HEU). A salient component of that initiative is the conversion of research reactors from HEU to low enriched uranium (LEU) fuels. An innovative fuel is being developed to replace HEU in high-power research reactors. The new LEU fuel is a monolithic fuel made from a U-Mo alloy foil encapsulated in Al-6061 cladding. In order to support the fuel qualification process, the Laser Shockwave Technique (LST) is being developed to characterize the clad-clad and fuel-clad interface strengths in fresh and irradiated fuel plates. LST is a non-contact method that uses lasers for the generation and detection of large amplitude acoustic waves to characterize interfaces in nuclear fuel plates. However, because the deposition of laser energy into the containment layer on a specimen's surface is intractably complex, the shock wave energy is inferred from the surface velocity measured on the backside of the fuel plate and the depth of the impression left on the surface by the high pressure plasma pulse created by the shock laser. To help quantify the stresses generated at the interfaces, a finite element method (FEM) model is being utilized. This paper will report on initial efforts to develop and validate the model by comparing numerical and experimental results for back surface velocities and front surface depressions in a single aluminum plate representative of the fuel cladding.
Constraining Hamiltonians from chiral effective field theory with neutron-star data
Multi-messenger observations of neutron stars (NSs) and their mergers have placed strong constraints on the dense-matter equation of state (EOS). The EOS, in turn, depends on microscopic nuclear interactions that are described by nuclear Hamiltonians. These Hamiltonians are commonly derived within chiral effective field theory (EFT). Ideally, multi-messenger observations of NSs could be used to directly inform our understanding of EFT interactions, but such a direct inference necessitates millions of model evaluations. This is computationally prohibitive because each evaluation requires us to calculate the EOS from a Hamiltonian by solving the quantum many-body problem with methods such as auxiliary-field diffusion Monte Carlo (AFDMC), which provides very accurate and precise solutions but at a significant computational cost. Additionally, we need to solve the stellar structure equations for each EOS which further slows down each model evaluation by a few seconds. In this work, we combine emulators for AFDMC calculations of neutron matter, built using parametric matrix models, and for the stellar structure equations, built using multilayer perceptron neural networks, with the PyCBC data-analysis framework to enable a direct inference of coupling constants in an EFT Hamiltonian using multi-messenger observations of NSs. We find that astrophysical data can provide informative constraints on two-nucleon couplings despite the high densities probed in NS interiors.
Solving the Orszag–Tang vortex magnetohydrodynamics problem with physics-constrained convolutional neural networks
We study the 2D Orszag–Tang vortex magnetohydrodynamics (MHD) problem through the use of physics-constrained convolutional neural networks (PCNNs) for forecasting the density, ρ, and the magnetic field, B, as well as the prediction of B given the velocity field v of the fluid. In addition to translation equivariance from the convolutional architecture, other physics constraints were embedded: absence of magnetic monopoles, non-negativity of ρ, use of only relevant variables, and the periodic boundary conditions of the problem. The use of only relevant variables and the hard constraint of non-negative ρ were found to facilitate learning greatly. The divergenceless condition ∇·B=0 was implemented as a hard constraint up to machine precision through the use of a magnetic potential to define B=∇×A. Residual networks and data augmentation were also used to improve performance. This allowed for some of the residual models to function as surrogate models and provide reasonably accurate simulations. For the prediction task, the PCNNs were evaluated against a physics-informed neural network, which had the ideal MHD induction equation as a soft constraint. Several models were able to generate highly accurate fields, which are visually almost indistinguishable and have low mean squared error. Only methods with built-in hard constraints produced physical fields with ∇·B=0. The use of PCNNs for MHD has the potential to produce physically consistent real-time simulations to serve as virtual diagnostics in cases where inferences must be made with limited observables.
Structure in the 3D Galaxy Distribution. III. Fourier Transforming the Universe: Phase and Power Spectra
We demonstrate the effectiveness of a relatively straightforward analysis of the complex 3D Fourier transform of galaxy coordinates derived from redshift surveys. Numerical demonstrations of this approach are carried out on a volume-limited sample of the Sloan Digital Sky Survey redshift survey. The direct unbinned transform yields a complex 3D data cube quite similar to that from the Fast Fourier Transform of finely binned galaxy positions. In both cases, deconvolution of the sampling window function yields estimates of the true transform. Simple power spectrum estimates from these transforms are roughly consistent with those using more elaborate methods. The complex Fourier transform characterizes spatial distributional properties beyond the power spectrum in a manner different from (and we argue is more easily interpreted than) the conventional multipoint hierarchy. We identify some threads of modern large-scale inference methodology that will presumably yield detections in new wider and deeper surveys.
Effect of mixing Ce{sup 3+} and Nd{sup 3+} ions in equimolar ratio on structural, optical and dielectric properties on pure cerium orthovanadate and neodymium orthovanadate
Highlights: • CeV, NdV and mixed CeNdV nanoparticle prepared by chemical co precipitation method. • With mixing of Ce{sup 3+} and Nd{sup 3+} morphology is totally changed in mixed CeNdV. • Optical band energy of CeV, NdV and CeNdV shows good photocatalyst under UV light. • Conduction mechanism in CeV due to large polaron and small polaron in CeNdV. - Abstract: Cerium orthovanadate, neodymium orthovanadate and mixed cerium neodymium orthovanadate nanoparticles was prepared by co-precipitation method. Powder X-ray diffraction reveals tetragonal zircon structure. Slight increase in lattice parameter, volume and decrease in X-ray density inferred that Ce{sup 3+} and Nd{sup 3+} ion replaces each other. Transmission electron microscopy suggests change in morphology with the effect of mixing and validates formation of nanocrystalline material. The infrared transmittance spectrum confirmed the presence of various functional groups. Dielectric properties as function of frequency show dielectric constant and loss tangent decreases with increase in frequency which is due to Maxwell–Wagner type interfacial polarization. The variation of AC conductivity measurement with frequency suggests conduction mechanism due to large polaron hopping in CeV whereas small polaron in mixed CeNdV. The activation energy decreases with rising frequency indicates the conduction mechanism is based on polaron hopping between localized states in disordered manner.
Improving River Routing Using a Differentiable Muskingum‐Cunge Model and Physics‐Informed Machine Learning
Recently, rainfall-runoff simulations in small headwater basins have been improved by methodological advances such as deep neural networks (NNs) and hybrid physics-NN models—particularly, a genre called differentiable modeling that intermingles NNs with physics to learn relationships between variables. However, hydrologic routing simulations, necessary for simulating floods in stem rivers downstream of large heterogeneous basins, had not yet benefited from these advances and it was unclear if the routing process could be improved via coupled NNs. We present a novel differentiable routing method (δMC-Juniata-hydroDL2) that mimics the classical Muskingum-Cunge routing model over a river network but embeds an NN to infer parameterizations for Manning's roughness (n) and channel geometries from raw reach-scale attributes like catchment areas and sinuosity. The NN was trained solely on downstream hydrographs. Synthetic experiments show that while the channel geometry parameter was unidentifiable, n can be identified with moderate precision. With real-world data, the trained differentiable routing model produced more accurate long-term routing results for both the training gage and untrained inner gages for larger subbasins (>2,000 km2) than either a machine learning model assuming homogeneity, or simply using the sum of runoff from subbasins. The n parameterization trained on short periods gave high performance in other periods, despite significant errors in runoff inputs. The learned n pattern was consistent with literature expectations, demonstrating the framework's potential for knowledge discovery, but the absolute values can vary depending on training periods. The trained n parameterization can be coupled with traditional models to improve national-scale hydrologic flood simulations.
A machine learning inversion scheme for determining interaction from scattering
Small angle scattering techniques have now been routinely used to quantitatively determine the potential of mean force in colloidal suspensions. However the numerical accuracy of data interpretation is often compounded by the approximations adopted by liquid state analytical theories. To circumvent this long standing issue, here we outline a machine learning strategy for determining the effective interaction in the condensed phases of matter using scattering. Via a case study of colloidal suspensions, we show that the effective potential can be probabilistically inferred from the scattering spectra without any restriction imposed by model assumptions. Comparisons to existing parametric approaches demonstrate the superior performance of this method in accuracy, efficiency, and applicability. This method can effectively enable quantification of interaction in highly correlated systems using scattering and diffraction experiments.
Fingerprints of hydrogen bonding in the terahertz dynamics of ethanol and water: An inelastic x-ray scattering study
Here, we used inelastic x-ray scattering methods to measure the terahertz spectrum of density fluctuations of ethanol in both liquid and solid phases. The results of a Bayesian inference-based lineshape analysis with a multiple excitation model and the comparison with a previous similar analysis on water indicate that the different structures induced by hydrogen bonds in ethanol and water have a profound influence on the respective dynamic responses, the latter being characterized by longer living and better resolved high-frequency acoustic excitations. In addition, we compare these findings with those obtained with an alternative approach based on the exponential expansion theory and ensuring sum rules fulfillment, demonstrating that the model’s choice directly impacts the number of spectral modes detected.
Bayesian inference of anisotropic 2D small-angle scattering from sparse measurement
Here, we present a Bayesian inference framework for reconstructing anisotropic two-dimensional small-angle scattering (2D SAS) patterns from sparse, noisy, or partially missing data. The method combines a symmetry-aware angular basis with radial Gaussian process priors to enable accurate, training-free interpolation and denoising. Computational benchmarks demonstrate reliable recovery of both isotropic and high-order anisotropic features under severe data reduction. Experimental validations on stretched polymers, sheared wormlike micelles, and carbon fibers show improved fidelity and resolution compared to raw measurements, achieving comparable accuracy with up to 50-fold fewer detected neutrons. This approach enables quantitative structural analysis under low-flux, time-limited, or single-shot conditions, extending the applicability of 2D SAS techniques to compact neutron sources and mechanically driven soft matter systems undergoing transient structural changes.
Application of Eye Tracking for Measurement and Evaluation in Human Factors Studies in Control Room Modernization
An important element of human factors engineering (HFE) pertains to measurement and evaluation (M&E). The role of HFE-M&E should be integrated throughout the entire control room modernization (CRM) process and be used for human-system performance evaluation and diagnostic purposes with resolving potential human engineering deficiencies (HEDs) and other human machine interface (HMI) design issues. NUREG-0711 describes how HFE in CRM should employ a hierarchical set of measures, particularly during integrated system validation (ISV), including plant performance, personnel task performance, situation awareness, cognitive workload, and anthropometric/ physiological factors. Historically, subjective measures have been primarily used since they are easier to collect and do not require specialized equipment. However, there are pitfalls with relying solely on subjective measures in M&E such that negatively impact reliability, sensitivity, and objectivity. Furthermore, as part of comprehensively capturing a diverse set of measures that strengthen findings and inferences made of the benefits from emerging technologies like advanced displays, this paper discusses the value of using eye tracking as an objective method that can be used in M&E. A brief description of eye tracking technology and relevant eye tracking measures is provided. Additionally, technical considerations and the unique challenges with using eye tracking in full-scaled simulations are addressed. Finally, this paper shares preliminary findings regarding the use of a wearable eye tracking system in a full-scale simulator study. These findings should help guide future full-scale simulator studies using eye tracking as a methodology to evaluate human-system performance.