Search NASA⌕ Search

SEARCH · Search NASA

Results for “Unstructured”

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

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

At least 1,279 records · Page 71

Clinical Natural Language Processing for Radiation Oncology: A Review and Practical Primer

Natural language processing (NLP), which aims to convert human language into expressions that can be analyzed by computers, is one of the most rapidly developing and widely used technologies in the field of artificial intelligence. Natural language processing algorithms convert unstructured free text data into structured data that can be extracted and analyzed at scale. In medicine, this unlocking of the rich, expressive data within clinical free text in electronic medical records will help untap the full potential of big data for research and clinical purposes. Recent major NLP algorithmic advances have significantly improved the performance of these algorithms, leading to a surge in academic and industry interest in developing tools to automate information extraction and phenotyping from clinical texts. Thus, these technologies are poised to transform medical research and alter clinical practices in the future. Radiation oncology stands to benefit from NLP algorithms if they are appropriately developed and deployed, as they may enable advances such as automated inclusion of radiation therapy details into cancer registries, discovery of novel insights about cancer care, and improved patient data curation and presentation at the point of care. However, challenges remain before the full value of NLP is realized, such as the plethora of jargon specific to radiation oncology, nonstandard nomenclature, a lack of publicly available labeled data for model development, and interoperability limitations between radiation oncology data silos. Successful development and implementation of high quality and high value NLP models for radiation oncology will require close collaboration between computer scientists and the radiation oncology community. Here, we present a primer on artificial intelligence algorithms in general and NLP algorithms in particular; provide guidance on how to assess the performance of such algorithms; review prior research on NLP algorithms for oncology; and describe future avenues for NLP in radiation oncology research and clinics.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Stratigraphic Identification with Airborne Electromagnetic Methods at the Hanford Site, Washington

Stratigraphic units can influence the fate and transport of subsurface contaminants within groundwater. Units having coarse-grained sediments act as preferential flow pathways, and therefore can accelerate the transport of contaminants to reach human and ecological receptors. At legacy waste sites, detailed knowledge of subsurface stratigraphy can be used for effective monitoring and remediation planning to help minimize risk to human health and the environment. Airborne electromagnetic (AEM) methods can non-invasively provide information on kilometer-scale or larger subsurface stratigraphic features and fill informational gaps in directly sampled data from sparsely located boreholes. In this paper, we present inversion results of a 412 line-km frequency-domain AEM survey to delineate subsurface stratigraphic features at the Hanford Site, located in southeastern Washington State. The inversion was performed using a massively parallel 3D electromagnetic modeling and inversion code, where the modeling is based on solving frequency-domain Maxwell’s equations using an unstructured-mesh finite-element method and the inversion employs a Gauss-Newton optimization scheme. The results are compared to an underlying geologic framework model (GFM), built by interpolating contact depths of stratigraphic units interpreted from site borehole datasets. In areas with good borehole coverage, the inversion results show a good match with the GFM to a depth of about 60 m. Outside of these areas, the inversion results exhibit inconsistencies from the assumptions made to create the GFM, demonstrating that the AEM survey results can be used to improve the understanding of the geological conceptual model.

47 OTHER INSTRUMENTATION↗

Modifying the resolving cysteine affects the structure and hydrogen peroxide reactivity of peroxiredoxin 2

Peroxiredoxin 2 (Prdx2) is a thiol peroxidase with an active site Cys (C52) that reacts rapidly with H 2 O 2 and other peroxides. The sulfenic acid product condenses with the resolving Cys (C172) to form a disulfide which is recycled by thioredoxin or GSH via mixed disulfide intermediates or undergoes hyperoxidation to the sulfinic acid. C172 lies near the C terminus, outside the active site. It is not established whether structural changes in this region, such as mixed disulfide formation, affect H 2 O 2 reactivity. To investigate, we designed mutants to cause minimal (C172S) or substantial (C172D and C172W) structural disruption. Stopped flow kinetics and mass spectrometry showed that mutation to Ser had minimal effect on rates of oxidation and hyperoxidation, whereas Asp and Trp decreased both by 100-fold. To relate to structural changes, we solved the crystal structures of reduced WT and C172S Prdx2. The WT structure is highly similar to that of the published hyperoxidized form. C172S is closely related but more flexible and as demonstrated by size exclusion chromatography and analytical ultracentrifugation, a weaker decamer. Size exclusion chromatography and analytical ultracentrifugation showed that the C172D and C172W mutants are also weaker decamers than WT, and small-angle X-ray scattering analysis indicated greater flexibility with partially unstructured regions consistent with C-terminal unfolding. We propose that these structural changes around C172 negatively impact the active site geometry to decrease reactivity with H 2 O 2 . This is relevant for Prdx turnover as intermediate mixed disulfides with C172 would also be disruptive and could potentially react with peroxides before resolution is complete.

59 BASIC BIOLOGICAL SCIENCES↗

Direct interaction of DNA repair protein tyrosyl DNA phosphodiesterase 1 and the DNA ligase III catalytic domain is regulated by phosphorylation of its flexible N-terminus

Tyrosyl DNA phosphodiesterase 1 (TDP1) and DNA Ligase IIIα (LigIIIα) are key enzymes in single-strand break (SSB) repair. TDP1 removes 3'-tyrosine residues remaining after degradation of DNA topoisomerase (TOP) 1 cleavage complexes trapped by either DNA lesions or TOP1 inhibitors. It is not known how TDP1 is linked to subsequent processing and LigIIIα-catalyzed joining of the SSB. Here we define a direct interaction between the TDP1 catalytic domain and the LigIII DNA-binding domain (DBD) regulated by conformational changes in the unstructured TDP1 N-terminal region induced by phosphorylation and/or alterations in amino acid sequence. Full-length and N-terminally truncated TDP1 are more effective at correcting SSB repair defects in TDP1 null cells compared with full-length TDP1 with amino acid substitutions of an N-terminal serine residue phosphorylated in response to DNA damage. TDP1 forms a stable complex with LigIII 170-755 , as well as full-length LigIIIα alone or in complex with the DNA repair scaffold protein XRCC1. Small-angle X-ray scattering and negative stain electron microscopy combined with mapping of the interacting regions identified a TDP1/LigIIIα compact dimer of heterodimers in which the two LigIII catalytic cores are positioned in the center, whereas the two TDP1 molecules are located at the edges of the core complex flanked by highly flexible regions that can interact with other repair proteins and SSBs. As TDP1and LigIIIα together repair adducts caused by TOP1 cancer chemotherapy inhibitors, the defined interaction architecture and regulation of this enzyme complex provide insights into a key repair pathway in nonmalignant and cancer cells.

59 BASIC BIOLOGICAL SCIENCES↗

Class imbalance in out-of-distribution datasets: Improving the robustness of the TextCNN for the classification of rare cancer types

In the last decade, the widespread adoption of electronic health record documentation has created huge opportunities for information mining. Natural language processing (NLP) techniques using machine and deep learning are becoming increasingly widespread for information extraction tasks from unstructured clinical notes. Disparities in performance when deploying machine learning models in the real world have recently received considerable attention. In the clinical NLP domain, the robustness of convolutional neural networks (CNNs) for classifying cancer pathology reports under natural distribution shifts remains understudied. In this research, we aim to quantify and improve the performance of the CNN for text classification on out-of-distribution (OOD) datasets resulting from the natural evolution of clinical text in pathology reports. We identified class imbalance due to different prevalence of cancer types as one of the sources of performance drop and analyzed the impact of previous methods for addressing class imbalance when deploying models in real-world domains. Our results show that our novel class-specialized ensemble technique outperforms other methods for the classification of rare cancer types in terms of macro F1 scores. We also found that traditional ensemble methods perform better in top classes, leading to higher micro F1 scores. Based on our findings, we formulate a series of recommendations for other ML practitioners on how to build robust models with extremely imbalanced datasets in biomedical NLP applications.

60 APPLIED LIFE SCIENCES↗

KEBLM: Knowledge-Enhanced Biomedical Language Models

Pretrained language models (PLMs) have demonstrated strong performance on many natural language processing (NLP) tasks. Despite their great success, these PLMs are typically pretrained only on unstructured free texts without leveraging existing structured knowledge bases that are readily available for many domains, especially scientific domains. As a result, these PLMs may not achieve satisfactory performance on knowledge-intensive tasks such as biomedical NLP. Comprehending a complex biomedical document without domain-specific knowledge is challenging, even for humans. Inspired by this observation, we propose a general framework for incorporating various types of domain knowledge from multiple sources into biomedical PLMs. We encode domain knowledge using lightweight adapter modules, bottleneck feed-forward networks that are inserted into different locations of a backbone PLM. For each knowledge source of interest, we pretrain an adapter module to capture the knowledge in a self-supervised way. We design a wide range of self-supervised objectives to accommodate diverse types of knowledge, ranging from entity relations to description sentences. Once a set of pretrained adapters is available, we employ fusion layers to combine the knowledge encoded within these adapters for downstream tasks. Each fusion layer is a parameterized mixer of the available trained adapters that can identify and activate the most useful adapters for a given input. Our method diverges from prior work by including a knowledge consolidation phase, during which we teach the fusion layers to effectively combine knowledge from both the original PLM and newly-acquired external knowledge using a large collection of unannotated texts. After the consolidation phase, the complete knowledge-enhanced model can be fine-tuned for any downstream task of interest to achieve optimal performance. Extensive experiments on many biomedical NLP datasets show that our proposed framework consistently improves the performance of the underlying PLMs on various downstream tasks such as natural language inference, question answering, and entity linking. These results demonstrate the benefits of using multiple sources of external knowledge to enhance PLMs and the effectiveness of the framework for incorporating knowledge into PLMs. Finally, while primarily focused on the biomedical domain in this work, our framework is highly adaptable and can be easily applied to other domains, such as the bioenergy sector.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

A piecewise linear finite element discretization of the diffusion equation for arbitrary polyhedral grids

We develop a piecewise linear (PWL) Galerkin finite element spatial discretization for the multi-dimensional radiation diffusion equation. It uses recently introduced piecewise linear weight and basis functions in the finite element approximation and it can be applied on arbitrary polygonal (2D) or polyhedral (3D) grids. We first demonstrate some analytical properties of the PWL method and perform a simple mode analysis to compare the PWL method with Palmer's vertex-centered finite-volume method and with a bilinear continuous finite element method. We then show that this new PWL method gives solutions comparable to those from Palmer's. However, since the PWL method produces a symmetric positive-definite coefficient matrix, it should be substantially more computationally efficient than Palmer's method, which produces an asymmetric matrix. We conclude that the Galerkin PWL method is an attractive option for solving diffusion equations on unstructured grids.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Electric and magnetic losses modeled by a stable hybrid with explicit-implicit time-stepping for Maxwell's equations

A stable hybridization of the finite-element method (FEM) and the finite-difference time-domain (FDTD) scheme for Maxwell's equations with electric and magnetic losses is presented for two-dimensional problems. The hybrid method combines the flexibility of the FEM with the efficiency of the FDTD scheme and it is based directly on Ampere's and Faraday's law. The electric and magnetic losses can be treated implicitly by the FEM on an unstructured mesh, which allows for local mesh refinement in order to resolve rapid variations in the material parameters and/or the electromagnetic field. It is also feasible to handle larger homogeneous regions with losses by the explicit FDTD scheme connected to an implicitly time-stepped and lossy FEM region. The hybrid method shows second-order convergence for smooth scatterers. The bistatic radar cross section (RCS) for a circular metal cylinder with a lossy coating converges to the analytical solution and an accuracy of 2% is achieved for about 20 points per wavelength. The monostatic RCS for an airfoil that features sharp corners yields a lower order of convergence and it is found to agree well with what can be expected for singular fields at the sharp corners. A careful convergence study with resolutions from 20 to 140 points per wavelength provides accurate extrapolated results for this non-trivial test case, which makes it possible to use as a reference problem for scattering codes that model both electric and magnetic losses.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Iterated upwind schemes for gas dynamics

A class of high-resolution schemes established in integration of anelastic equations is extended to fully compressible flows, and documented for unsteady (and steady) problems through a span of Mach numbers from zero to supersonic. The schemes stem from iterated upwind technology of the multidimensional positive definite advection transport algorithm (MPDATA). The derived algorithms employ standard and modified forms of the equations of gas dynamics for conservation of mass, momentum and either total or internal energy as well as potential temperature. Numerical examples from elementary wave propagation, through computational aerodynamics benchmarks, to atmospheric small- and large-amplitude acoustics with intricate wave-flow interactions verify the approach for both structured and unstructured meshes, and demonstrate its flexibility and robustness.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

A priori mesh quality estimation via direct relation between truncation error and mesh distortion

The purpose of the present work is the derivation and evaluation of a priori mesh quality indicators for structured, unstructured, as well as hybrid grids. Emphasis is placed on deriving direct relations between the indicators and mesh distortion. The work is based on use of the finite volume discretization for evaluation of first order spatial derivatives. The analytic form of the truncation error is derived and applied to elementary types of mesh distortion including typical hybrid grid interfaces. The corresponding analytic expressions provide direct relations between computational accuracy and the degree of stretching, skewness, shearing and non-alignment of the mesh.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Multicloud: Multigrid convergence with a meshless operator

The primary objective of this work is to develop and test a new convergence acceleration technique we call multicloud. Multicloud is well-founded in the mathematical basis of multigrid, but relies on a meshless operator on coarse levels. The meshless operator enables extremely simple and automatic coarsening procedures for arbitrary meshes using arbitrary fine level discretization schemes. The performance of multicloud is compared with established multigrid techniques for structured and unstructured meshes for the Euler equations on two-dimensional test cases. Results indicate comparable convergence rates per unit work for multicloud and multigrid. However, because of its mesh and scheme transparency, multicloud may be applied to a wide array of problems with no modification of fine level schemes as is often required with agglomeration techniques. The implication is that multicloud can be implemented in a completely modular fashion, allowing researchers to develop fine level algorithms independent of the convergence accelerator for complex three-dimensional problems.

97 MATHEMATICS AND COMPUTING↗

An adaptive algorithm for simulation of stochastic reaction-diffusion processes

We propose an adaptive hybrid method suitable for stochastic simulation of diffusion dominated reaction-diffusion processes. For such systems, simulation of the diffusion requires the predominant part of the computing time. In order to reduce the computational work, the diffusion in parts of the domain is treated macroscopically, in other parts with the tau-leap method and in the remaining parts with Gillespie's stochastic simulation algorithm (SSA) as implemented in the next subvolume method (NSM). The chemical reactions are handled by SSA everywhere in the computational domain. A trajectory of the process is advanced in time by an operator splitting technique and the timesteps are chosen adaptively. The spatial adaptation is based on estimates of the errors in the tau-leap method and the macroscopic diffusion. The accuracy and efficiency of the method are demonstrated in examples from molecular biology where the domain is discretized by unstructured meshes.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Scalable parallel methods for monolithic coupling in fluid-structure interaction with application to blood flow modeling

We introduce and study numerically a scalable parallel finite element solver for the simulation of blood flow in compliant arteries. The incompressible Navier-Stokes equations are used to model the fluid and coupled to an incompressible linear elastic model for the blood vessel walls. Our method features an unstructured dynamic mesh capable of modeling complicated geometries, an arbitrary Lagrangian-Eulerian framework that allows for large displacements of the moving fluid domain, monolithic coupling between the fluid and structure equations, and fully implicit time discretization. Simulations based on blood vessel geometries derived from patient-specific clinical data are performed on large supercomputers using scalable Newton-Krylov algorithms preconditioned with an overlapping restricted additive Schwarz method that preconditions the entire fluid-structure system together. The algorithm is shown to be robust and scalable for a variety of physical parameters, scaling to hundreds of processors and millions of unknowns.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Adaptive spacetime method using Riemann jump conditions for coupled atomistic-continuum dynamics

We combine the Spacetime Discontinuous Galerkin (SDG) method for elastodynamics with the mathematically consistent Atomistic Discontinuous Galerkin (ADG) method in a new scheme that concurrently couples continuum and atomistic models of dynamic response in solids. The formulation couples non-overlapping continuum and atomistic models across sharp interfaces by weakly enforcing jump conditions, for both momentum balance and kinematic compatibility, using Riemann values to preserve the characteristic structure of the underlying hyperbolic system. Momentum balances to within machine-precision accuracy over every element, on each atom, and over the coupled system, with small, controllable energy dissipation in the continuum region that ensures numerical stability. When implemented on suitable unstructured spacetime grids, the continuum SDG model offers linear computational complexity in the number of elements and powerful adaptive analysis capabilities that readily bridge between atomic and continuum scales in both space and time. A special trace operator for the atomic velocities and an associated atomistic traction field enter the jump conditions at the coupling interface. The trace operator depends on parameters that specify, at the scale of the atomic spacing, the position of the coupling interface relative to the atoms. In a key finding, we demonstrate that optimizing these parameters suppresses spurious reflections at the coupling interface without the use of non-physical damping or special boundary conditions. We formulate the implicit SDG-ADG coupling scheme in up to three spatial dimensions, and describe an efficient iterative solution scheme that outperforms common explicit schemes, such as the Velocity Verlet integrator. Numerical examples, in 1dxtime and employing both linear and nonlinear potentials, demonstrate the performance of the SDG-ADG method and show how adaptive spacetime meshing reconciles disparate time steps and resolves atomic-scale signals in the continuum.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Towards a Scalable Fully-Implicit Fully-coupled Resistive MHD Formulation with Stabilized FE Methods

This paper explores the development of a scalable, nonlinear, fully-implicit stabilized unstructured finite element (FE) capability for 2D incompressible (reduced) resistive MHD. The discussion considers the implementation of a stabilized FE formulation in context of a fully-implicit time integration and direct-to-steady-state solution capability. The nonlinear solver strategy employs Newton-Krylov methods, which are preconditioned using fully-coupled algebraic multilevel preconditioners. These preconditioners are shown to enable a robust, scalable and efficient solution approach for the large-scale sparse linear systems generated by the Newton linearization. Verification results demonstrate the expected order-of-accuracy for the stabilized FE discretization. The approach is tested on a variety of prototype problems, including both low-Lundquist number (e.g., an MHD Faraday conduction pump and a hydromagnetic Rayleigh-Bernard linear stability calculation) and moderately-high Lundquist number (magnetic island coalescence problem) examples. Initial results that explore the scaling of the solution methods are presented on up to 4096 processors for problems with up to 64M unknowns on a CrayXT3/4. Additionally, a large-scale proof-of-capability calculation for 1 billion unknowns for the MHD Faraday pump problem on 24,000 cores is presented.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

High-order finite-element seismic wave propagation modeling with MPI on a large GPU cluster

We implement a high-order finite-element application, which performs the numerical simulation of seismic wave propagation resulting for instance from earthquakes at the scale of a continent or from active seismic acquisition experiments in the oil industry, on a large cluster of NVIDIA Tesla graphics cards using the CUDA programming environment and non-blocking message passing based on MPI. Contrary to many finite-element implementations, ours is implemented successfully in single precision, maximizing the performance of current generation GPUs. We discuss the implementation and optimization of the code and compare it to an existing very optimized implementation in C language and MPI on a classical cluster of CPU nodes. We use mesh coloring to efficiently handle summation operations over degrees of freedom on an unstructured mesh, and non-blocking MPI messages in order to overlap the communications across the network and the data transfer to and from the device via PCIe with calculations on the GPU. We perform a number of numerical tests to validate the single-precision CUDA and MPI implementation and assess its accuracy. We then analyze performance measurements and depending on how the problem is mapped to the reference CPU cluster, we obtain a speedup of 20x or 12x.

97 MATHEMATICS AND COMPUTING↗

Implicit high-order method for calculating rarefied gas flow in a planar microchannel

An efficient numerical algorithm for calculating rarefied gas flows in planar microchannels on the basis of the Boltzmann kinetic equation with the linearized S-model collision integral is presented. The algorithm consists of a high-order spatial discretisation on unstructured meshes, conservative procedure to calculate macroscopic quantities and efficient one-step implicit time evolution method. It therefore works across all flow regimes from the free-molecular to nearly continuum ones and provides rapid convergence to steady state solutions. The parallel implementation is provided via MPI programming paradigm. The performance of the method is illustrated by computing the flow for length to the width ratios for up to 10{sup 3} in the wide range of Knudsen numbers. Numerical estimates of efficiency of implicit time marching and parallel performance as well as convergence studies in both physical and velocity spaces are provided. The end effects and other essentially two-dimensional features of the flow are analysed and a detailed comparison with the case of an infinitely long channel is carried out. The presented numerical data can be used as reference for future studies of nonlinear flows as well as for comparison with other numerical approaches and flow models.

97 MATHEMATICS AND COMPUTING↗

An efficient parallel implementation of explicit multirate Runge–Kutta schemes for discontinuous Galerkin computations

Although explicit time integration schemes require small computational efforts per time step, their efficiency is severely restricted by their stability limits. Indeed, the multi-scale nature of some physical processes combined with highly unstructured meshes can lead some elements to impose a severely small stable time step for a global problem. Multirate methods offer a way to increase the global efficiency by gathering grid cells in appropriate groups under local stability conditions. These methods are well suited to the discontinuous Galerkin framework. The parallelization of the multirate strategy is challenging because grid cells have different workloads. The computational cost is different for each sub-time step depending on the elements involved and a classical partitioning strategy is not adequate any more. In this paper, we propose a solution that makes use of multi-constraint mesh partitioning. It tends to minimize the inter-processor communications, while ensuring that the workload is almost equally shared by every computer core at every stage of the algorithm. Particular attention is given to the simplicity of the parallel multirate algorithm while minimizing computational and communication overheads. Our implementation makes use of the MeTiS library for mesh partitioning and the Message Passing Interface for inter-processor communication. Performance analyses for two and three-dimensional practical applications confirm that multirate methods preserve important computational advantages of explicit methods up to a significant number of processors.

97 MATHEMATICS AND COMPUTING↗