Search NASA⌕ Search

SEARCH · Search NASA

Results for “kernel method”

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 415 records · Page 23

Data Driven Optimization Framework for Volt-age Regulation in Distribution Systems

This letter proposes a data-driven optimization framework for voltage regulation problems to address the challenge of model inaccuracy and parameter varying. To achieve online voltage optimization, the recursive kernel regression and interior point methods are integrated. The IEEE 123-Bus system and EPRI Ckt5 feeder are selected to validate the effectiveness of the proposed data-driven optimization framework. The proposed method is also compared with a linear function based method.

Hong, Tianqi↗

Blueprints for Training Information Bottlenecks for Collider Analyses

Dimensionality reduction is a crucial aspect of data analysis in high energy physics, even if accompanied by information loss. Several methods, including histogram- and kernel-based analyses, are only computationally feasible for low-dimensional data. Furthermore, simulation models used in HEP can often only be validated for low-dimensional data. We provide several blueprints for using machine learning to create low-dimensional data representations (continuous event variables and discrete classification labels) for use in signal discovery and parameter estimation tasks. We also describe how to design the learned representation to facilitate a) searches with unknown model parameters and b) validation of simulation models in data control regions.

43 PARTICLE ACCELERATORS↗

Echo mapping of active galactic nuclei broad-line regions: Fundamental algorithms

We formulate and test a series of algorithms for echo mapping the emission-line regions near active galactic nuclei from measurements of correlated variability in their line and continuum light curves. The linear regularization method (LRM) employs a direct inversion of evenly spaced light-curve data, with a regularization parameter that can be used to control the trade-off between noise and resolution. Matrix formulas express the formal solution as well as its variance and covariance in terms of uncertainties in the measurements. Unlike the maximum-entropy method (MEM), LRM applies to kernels with both positive and negative values, but the results are somewhat limited by ringing effects. A positivity constraint proves effective in controlling the ringing. MEM combines regularization and positivity in a natural way, but similar results are also found using positivity constraints with nonentropic regularization functions. Direct inversions of unevenly sampled light curves require interpolating the noisy data. In this case better results are found by solving for both the continuum light curve and kernel function in a simultaneous fit to the data. Our conclusion is that while echo mapping currently gives ambiguous results, the algorithms are not the limiting factor. Progress depends on efforts to increase the accuracy and completeness of sampling of the observed light curves.

Vio, Roberto↗

A reproducing kernel enhanced approach for peridynamic solutions

The most common discretization method for peridynamic models used in engineering problems is the node-based meshfree approach. This method discretizes peridynamic domains by a set of nodes, each associated with a nodal cell with a characteristic volume, leading to a particle-based description of continuum systems. The behavior of each particle is then considered representative of its cell. This limits the convergence rate to the first order. In this paper, we introduce a reproducing kernel (RK) approximation to the field variables in the peridynamic equations to increase the order of convergence of peridynamic numerical solutions. Furthermore, the numerical results demonstrate improved convergence rates in static peridynamic problems using the proposed method.

97 MATHEMATICS AND COMPUTING↗

Statistical Sampling Plan for AGR-5/6/7 Fuel Materials

Kernels, coated particles, and compacts will be manufactured to predefined specifications for irradiation in the AGR 5/6/7 experiments. Many tests will be performed for quality assurance and most of these tests correspond to criteria that have to be met with 95% confidence, based on random samples. This report describes the statistical acceptance testing methods and gives minimum sample sizes for kernels to achieve 95% confidence. Coated particles and compacts will be addressed in revisions to this report. The methods used for sample size determination are described in detail in INL/EXT 05 00349, Statistical Methods Handbook for Advanced Gas Reactor Fuel Materials (Einerson 2005). The statistical approach is to collect a representative sample, apply an acceptance test to the sample, and accept or reject the product based on this test for each criterion. The criteria or “product specifications” are contained in SPC-1352, “AGR 5/6/7 Fuel Specification,” Revision 1 and are duplicated in this plan.

Approaches to reducing photon dose calculation errors near metal implants

Purpose: Dose calculation errors near metal implants are caused by limitations of the dose calculation algorithm in modeling tissue/metal interface effects as well as density assignment errors caused by imaging artifacts. The purpose of this study was to investigate two strategies for reducing dose calculation errors near metal implants: implementation of metal-based energy deposition kernels in the convolution/superposition (C/S) dose calculation method and use of metal artifact reduction methods for computed tomography (CT) imaging. Methods: Both error reduction strategies were investigated using a simple geometric slab phantom with a rectangular metal insert (composed of titanium or Cerrobend), as well as two anthropomorphic phantoms (one with spinal hardware and one with dental fillings), designed to mimic relevant clinical scenarios. To assess the dosimetric impact of metal kernels, the authors implemented titanium and silver kernels in a commercial collapsed cone C/S algorithm. To assess the impact of CT metal artifact reduction methods, the authors performed dose calculations using baseline imaging techniques (uncorrected 120 kVp imaging) and three commercial metal artifact reduction methods: Philips Healthcare’s O-MAR, GE Healthcare’s monochromatic gemstone spectral imaging (GSI) using dual-energy CT, and GSI with metal artifact reduction software (MARS) applied. For the simple geometric phantom, radiochromic film was used to measure dose upstream and downstream of metal inserts. For the anthropomorphic phantoms, ion chambers and radiochromic film were used to quantify the benefit of the error reduction strategies. Results: Metal kernels did not universally improve accuracy but rather resulted in better accuracy upstream of metal implants and decreased accuracy directly downstream. For the clinical cases (spinal hardware and dental fillings), metal kernels had very little impact on the dose calculation accuracy (<1.0%). Of the commercial CT artifact reduction methods investigated, the authors found that O-MAR was the most consistent method, resulting in either improved dose calculation accuracy (dental case) or little impact on calculation accuracy (spine case). GSI was unsuccessful at reducing the severe artifacts caused by dental fillings and had very little impact on calculation accuracy. GSI with MARS on the other hand gave mixed results, sometimes introducing metal distortion and increasing calculation errors (titanium rectangular implant and titanium spinal hardware) but other times very successfully reducing artifacts (Cerrobend rectangular implant and dental fillings). Conclusions: Though successful at improving dose calculation accuracy upstream of metal implants, metal kernels were not found to substantially improve accuracy for clinical cases. Of the commercial artifact reduction methods investigated, O-MAR was found to be the most consistent candidate for all-purpose CT simulation imaging. The MARS algorithm for GSI should be used with caution for titanium implants, larger implants, and implants located near heterogeneities as it can distort the size and shape of implants and increase calculation errors.

60 APPLIED LIFE SCIENCES↗

On the interpretation of kernels - Computer simulation of responses to impulse pairs

A method is presented for the use of a unit impulse response and responses to impulse pairs of variable separation in the calculation of the second-degree kernels of a quadratic system. A quadratic system may be built from simple linear terms of known dynamics and a multiplier. Computer simulation results on quadratic systems with building elements of various time constants indicate reasonably that the larger time constant term before multiplication dominates in the envelope of the off-diagonal kernel curves as these move perpendicular to and away from the main diagonal. The smaller time constant term before multiplication combines with the effect of the time constant after multiplication to dominate in the kernel curves in the direction of the second-degree impulse response, i.e., parallel to the main diagonal. Such types of insight may be helpful in recognizing essential aspects of (second-degree) kernels; they may be used in simplifying the model structure and, perhaps, add to the physical/physiological understanding of the underlying processes.

Hung, G.↗

Technical Note: Image filtering to make computer-aided detection robust to image reconstruction kernel choice in lung cancer CT screening

Purpose: In lung cancer computed tomography (CT) screening, the performance of a computer-aided detection (CAD) system depends on the selection of the image reconstruction kernel. To reduce this dependence on reconstruction kernels, the authors propose a novel application of an image filtering method previously proposed by their group. Methods: The proposed filtering process uses the ratio of modulation transfer functions (MTFs) of two reconstruction kernels as a filtering function in the spatial-frequency domain. This method is referred to as MTF{sub ratio} filtering. Test image data were obtained from CT screening scans of 67 subjects who each had one nodule. Images were reconstructed using two kernels: f{sub STD} (for standard lung imaging) and f{sub SHARP} (for sharp edge-enhancement lung imaging). The MTF{sub ratio} filtering was implemented using the MTFs measured for those kernels and was applied to the reconstructed f{sub SHARP} images to obtain images that were similar to the f{sub STD} images. A mean filter and a median filter were applied (separately) for comparison. All reconstructed and filtered images were processed using their prototype CAD system. Results: The MTF{sub ratio} filtered images showed excellent agreement with the f{sub STD} images. The standard deviation for the difference between these images was very small, ∼6.0 Hounsfield units (HU). However, the mean and median filtered images showed larger differences of ∼48.1 and ∼57.9 HU from the f{sub STD} images, respectively. The free-response receiver operating characteristic (FROC) curve for the f{sub SHARP} images indicated poorer performance compared with the FROC curve for the f{sub STD} images. The FROC curve for the MTF{sub ratio} filtered images was equivalent to the curve for the f{sub STD} images. However, this similarity was not achieved by using the mean filter or median filter. Conclusions: The accuracy of MTF{sub ratio} image filtering was verified and the method was demonstrated to be effective for reducing the kernel dependence of CAD performance.

60 APPLIED LIFE SCIENCES↗

Multilevel summation methods for efficient evaluation of long-range pairwise interactions in atomistic and coarse-grained molecular simulation.

The availability of efficient algorithms for long-range pairwise interactions is central to the success of numerous applications, ranging in scale from atomic-level modeling of materials to astrophysics. This report focuses on the implementation and analysis of the multilevel summation method for approximating long-range pairwise interactions. The computational cost of the multilevel summation method is proportional to the number of particles, N, which is an improvement over FFTbased methods who s cost is asymptotically proportional to N logN. In addition to approximating electrostatic forces, the multilevel summation method can be use to efficiently approximate convolutions with long-range kernels. As an application, we apply the multilevel summation method to a discretized integral equation formulation of the regularized generalized Poisson equation. Numerical results are presented using an implementation of the multilevel summation method in the LAMMPS software package. Preliminary results show that the computational cost of the method scales as expected, but there is still a need for further optimization.

Bond, Stephen D.↗

SU-F-SPS-09: Parallel MC Kernel Calculations for VMAT Plan Improvement

Purpose: Adding kernels (small perturbations in leaf positions) to the existing apertures of VMAT control points may improve plan quality. We investigate the calculation of kernel doses using a parallelized Monte Carlo (MC) method. Methods: A clinical prostate VMAT DICOM plan was exported from Eclipse. An arbitrary control point and leaf were chosen, and a modified MLC file was created, corresponding to the leaf position offset by 0.5cm. The additional dose produced by this 0.5 cm × 0.5 cm kernel was calculated using the DOSXYZnrc component module of BEAMnrc. A range of particle history counts were run (varying from 3 × 10{sup 6} to 3 × 10{sup 7}); each job was split among 1, 10, or 100 parallel processes. A particle count of 3 × 10{sup 6} was established as the lower range because it provided the minimal accuracy level. Results: As expected, an increase in particle counts linearly increases run time. For the lowest particle count, the time varied from 30 hours for the single-processor run, to 0.30 hours for the 100-processor run. Conclusion: Parallel processing of MC calculations in the EGS framework significantly decreases time necessary for each kernel dose calculation. Particle counts lower than 1 × 10{sup 6} have too large of an error to output accurate dose for a Monte Carlo kernel calculation. Future work will investigate increasing the number of parallel processes and optimizing run times for multiple kernel calculations.

60 APPLIED LIFE SCIENCES↗

An adaptive fast multipole method in Cartesian basis, enabled by algorithmic differentiation

A novel form of the fast multipole method is realized by the combination of adaptive data structures with implementation of the multipole expansions and their translations in Cartesian basis, and enabled by algorithmic differentiation. It may be possible to generalize it to any kernel that allows its formulation as a function of small quantities with respect to which they can be expanded. There is no hard and fast rule recipe that works for all cases. However, it is shown how to do it for the softened Coulomb kernel, which is different from the Coulomb kernel. The characteristics and performance of the method are mapped and shown to provide an attractive option for a generic fast multipole algorithm in a wide region of the parameter space. The algorithm is illustrated with a beam dynamics application.

97 MATHEMATICS AND COMPUTING↗

TU-CD-304-07: Intensity Modulated Electron Beam Therapy Employing Small Fields in Virtual Scanning Mode

Purpose: Dynamic electron radiation therapies such as dynamic electron arc radiotherapy (DEAR) utilize small fields to provide target conformity and fluence modulation. The purpose of this study is to demonstrate the feasibility of virtual scanning mode using small fields. Methods: Monte Carlo simulations (EGSnrc/BEAMnrc/DOSXYZnrc) were performed using validated Varian TrueBeam phase space files for electron beam energies of 6, 9, 12, and 16 MeV and square/circular fields (1×1/1, 2×2/2, 3×3/3, 4×4/4, 5×5/5 cm{sup 2}/cm diameter). Resulting dose distributions (kernels) were used for subsequent calculations. The following analyses were performed: (1) Comparison of composite square fields and reference 10×10 cm{sup 2} dose distributions and (2) Scanning beam deliveries for square and circular fields realized as the convolution of kernels and scanning pattern. Preliminary beam weight and pattern optimization were also performed. Two linear scans of 10 cm with/without overlap were modeled. Comparison metrics included depth and orthogonal profiles at dmax. Results: (1) Composite fields regained reference depth dose profiles for most energies and fields within 5%. Smaller kernels and higher energies increased dose in the build-up and Bremsstrahlung region (30%, 16MeV and 1×1 cm{sup 2}), while reference dmax was maintained for all energies and composite fields. Smaller kernels (<2×2 cm{sup 2}) maintained penumbra and field size within 0.2 cm, and flatness within 2%. Deterioration of penumbra for larger kernels (5×5 cm{sup 2}) were observed. Balancing desirable dosimetry and efficiencies suggests that smaller kernels are used at edges and larger kernels in the center of the target. (2) Beam weight optimization improved cross-plane penumbra (0.2 cm) and increased the field size (0.4 cm) on average. In-plane penumbra and field size remained unchanged. Overlap depended on kernel size and optimal overlap resulted in flatness ±2%. Conclusion: Dynamic electron beam therapy in virtual scanning mode is feasible by employing small fields to achieve desired dose distributions and acceptable efficiencies.

60 APPLIED LIFE SCIENCES↗

Fast Bayesian blind deconvolution with Huber Super Gaussian priors

We report expectation Maximization (EM) based inference has already proven to be a very powerful tool to solve blind image deconvolution (BID) problems. Unfortunately, three important problems still impede the application of EM in BID: the undesirable saddle points and local minima caused by highly nonconvex priors, the instability around zero of some of the most interesting sparsity promoting priors, and the intrinsic high computational cost of the corresponding BID algorithm. In this paper we first show how Super Gaussian priors can be made numerically tractable around zero by introducing the family of Huber Super Gaussian priors and then present a fast EM based blind deconvolution method formulated in the image space. In the proposed computational approach, image and kernel estimation are performed by using the Alternating Direction Method of Multipliers (ADMM), which allows to exploit the advantages of FFT computation. For highly nonconvex priors, we propose a Smooth ADMM (SADMM) approach to avoid poor BID estimates. In conclusion, extensive experiments demonstrate that the proposed method significantly outperforms state-of-the-art BID methods in terms of quality of the reconstructions and speed.

97 MATHEMATICS AND COMPUTING↗

Acceleration of the Parameterization of Unified Microphysics Across Scales (PUMAS) on the Graphics Processing Unit (GPU) With Directive-Based Methods

Cloud microphysics is one of the most time-consuming components in a climate model. In this study, we port the cloud microphysics parameterization in the Community Atmosphere Model (CAM), known as Parameterization of Unified Microphysics Across Scales (PUMAS), from CPU to GPU to seek a computational speedup. The directive-based methods (OpenACC and OpenMP target offload) are determined as the best fit specifically for our development practices, which enable a single version of source code to run either on the CPU or GPU, and yield a better portability and maintainability. Their performance is first examined in a PUMAS stand-alone kernel and the directive-based methods can outperform a CPU node as long as there is enough computational burden on the GPU. A consistent behavior is observed when we run PUMAS on the GPU in a practical CAM simulation. A 3.6× speedup of the PUMAS execution time, including data movement between CPU and GPU, is achieved at a coarse horizontal resolution (8 NVIDIA V100 GPUs against 36 Intel Skylake CPU cores). This speedup further increases up to 5.4× at a high resolution (24 NVIDIA V100 GPUs against 108 Intel Skylake CPU cores), which highlights the fact that GPU favors larger problem size. This study demonstrates that using GPU in a CAM simulation can save noticeable computational costs even with a small portion of code being GPU-enabled. Therefore, we are encouraged to port more parameterizations to GPU to take advantage of its computational benefit.

54 ENVIRONMENTAL SCIENCES↗

Convolution-controlled rotation and scale invariance in optical correlation

A method is presented for evoking a controlled, continuously-variable degree of rotation- and scale-invariance in optical correlation; the method is suitable for off-line computation of filters, though not for real-time computation. While a closed-form solution for the blur kernels has thus far evaded solution, a digital approximation method has been presented. A simulated correlation run with real, frame-grabbed imagery has indicated the method's desired performance. These Gaussian blur kernels can be replaced with box-car kernels or other blur kernels suitable for the given correlation-task.

Juday, Richard D.↗

Performance-Portable Particle Advection with VTK-m

Particle advection is the fundamental kernel behind most vector field visualization methods. Yet, the efficient parallel computation of large amounts of particle traces remains challenging. This is exacerbated by the variety of hardware trends in today's HPC arena, including increasing core counts in classical CPUs, many-core designs such as the Intel Xeon Phi, and massively parallel GPUs. The dedicated optimization of a particle advection kernel for each individual target architecture is both time-consuming and error prone. In this paper, we propose a performance-portable algorithm for particle advection. Our algorithm is based on the recently introduced VTK-m system and chiefly relies on its device adapter abstraction. We demonstrate the general portability of our implementation across a wide variety of hardware. Finally, our evaluation shows that our hardware-agnostic algorithm has comparable performance to hardware-specific algorithms.

Pugmire, Dave↗

Compactly‐Supported Nonstationary Kernels for Computing Exact Gaussian Processes on Big Data

The Gaussian process (GP) is a widely used method for analyzing large-scale data sets, including spatio-temporal measurements of nonlinear processes that are now commonplace in the environmental sciences. Traditional implementations of GPs involve stationary kernels (also termed covariance functions) that limit their flexibility, and exact methods for inference that prevent application to data sets with more than about 10,000 points. Modern approaches to address stationarity assumptions generally fail to accommodate large data sets, while all attempts to address scalability focus on approximating the Gaussian likelihood, which can involve subjectivity and lead to inaccuracies. In this work, we explicitly derive an alternative kernel that can discover and encode both sparsity and nonstationarity. We embed the kernel within a fully Bayesian GP model and leverage high-performance computing resources to enable the analysis of massive data sets. We demonstrate the favorable performance of our novel kernel relative to existing exact and approximate GP methods across a variety of synthetic data examples. Furthermore, we conduct space–time prediction based on more than 1 million measurements of daily maximum temperature and verify that our results outperform state-of-the-art methods in the Earth sciences. More broadly, having access to exact GPs that use ultra-scalable, sparsity-discovering, nonstationary kernels allows GP methods to truly compete with a wide variety of machine learning methods.

Gaussian processes↗

Parametric Deformation of Discrete Geometry for Aerodynamic Shape Design

We present a versatile discrete geometry manipulation platform for aerospace vehicle shape optimization. The platform is based on the geometry kernel of an open-source modeling tool called Blender and offers access to four parametric deformation techniques: lattice, cage-based, skeletal, and direct manipulation. Custom deformation methods are implemented as plugins, and the kernel is controlled through a scripting interface. Surface sensitivities are provided to support gradient-based optimization. The platform architecture allows the use of geometry pipelines, where multiple modelers are used in sequence, enabling manipulation difficult or impossible to achieve with a constructive modeler or deformer alone. We implement an intuitive custom deformation method in which a set of surface points serve as the design variables and user-specified constraints are intrinsically satisfied. We test our geometry platform on several design examples using an aerodynamic design framework based on Cartesian grids. We examine inverse airfoil design and shape matching and perform lift-constrained drag minimization on an airfoil with thickness constraints. A transport wing-fuselage integration problem demonstrates the approach in 3D. In a final example, our platform is pipelined with a constructive modeler to parabolically sweep a wingtip while applying a 1-G loading deformation across the wingspan. This work is an important first step towards the larger goal of leveraging the investment of the graphics industry to improve the state-of-the-art in aerospace geometry tools.

Anderson, George R.↗