Search NASA⌕ Search

SEARCH · Search NASA

Results for “convolution”

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 613 records · Page 34

Physics-Driven Convolutional Autoencoder Approach for CFD Data Compressions: Preprint

With the growing size and complexity of turbulent flow models, data compression approaches are of the utmost importance to analyze, visualize, or restart the simulations. Recently, in-situ autoencoder-based compression approaches have been proposed and shown to be effective at producing reduced representations of turbulent flow data. However, these approaches focus solely on training the model using point-wise sample reconstruction losses that do not take advantage of the physical properties of turbulent flows. In this paper, we show that training autoencoders with additional physics-informed regularizations, e.g., enforcing incompressibility and preserving enstrophy, improves the compression model in three ways: (i) the compressed data better conform to known physics for homogeneous isotropic turbulence without negatively impacting point-wise reconstruction quality, (ii) inspection of the gradients of the trained model uncovers changes to the learned compression mapping that can facilitate the use of explainability techniques, and (iii) as a performance byproduct, training losses are shown to converge up to 12x faster than the baseline model.

auto-encoders↗

3D segmentation using space carving and 2D convolutional neural networks

A system for generating a 3D segmentation of a target volume is provided. The system accesses views of an X-ray scan of a target volume. The system applies a 2D CNN to each view to generate a 2D multi-channel feature vector for each view. The system applies a space carver to generate a 3D channel volume for each channel based on the 2D multi-channel feature vectors. The system then applies a linear combining technique to the 3D channel volumes to generate a 3D multi-label map that represents a 3D segmentation of the target volume.

97 MATHEMATICS AND COMPUTING↗

Accelerated Cartesian expansion (ACE) based framework for the rapid evaluation of diffusion, lossy wave, and Klein-Gordon potentials

Diffusion, lossy wave, and Klein–Gordon equations find numerous applications in practical problems across a range of diverse disciplines. The temporal dependence of all three Green’s functions are characterized by an infinite tail. This implies that the cost complexity of the spatio-temporal convolutions, associated with evaluating the potentials, scales as O(N s 2 N t 2 ), where N s and N t are the number of spatial and temporal degrees of freedom, respectively. In this paper, we discuss two new methods to rapidly evaluate these spatio-temporal convolutions by exploiting their block-Toeplitz nature within the framework of accelerated Cartesian expansions (ACE). The first scheme identifies a convolution relation in time amongst ACE harmonics and the fast Fourier transform (FFT) is used for efficient evaluation of these convolutions. The second method exploits the rank deficiency of the ACE translation operators with respect to time and develops a recursive numerical compression scheme for the efficient representation and evaluation of temporal convolutions. It is shown that the cost of both methods scales as O(N s N t log 2 N t ). Furthermore, several numerical results are presented for the diffusion equation to validate the accuracy and efficacy of the fast algorithms developed here.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

A Discrete Hankel Transform Approach to Nuclear Data Processing for Fusion Applications

This study introduces advancements to the numerical solutions employed in the processing of nuclear data for fusion applications. It leverages the convolution theorem and Fourier transform techniques to enhance computational efficiency and broaden applicability. Building upon a previously reported discrete Hankel transform approach for Doppler broadening, this work refines the solution of convolution integrals central to these applications. The methodology provides a general and unified framework for evaluating any convolution operation, regardless of whether the underlying problem involves temperature effects in nuclear reactions. The applicability to the nuclear data processing for fusion is demonstrated by deriving the convolution integrals for some of the fusion-related quantities. As before, the convolution operation utilizes a Gaussian-based kernel; however, the discrete Hankel transform of order $𝛼$ = $\frac{1}{2}$ is now applied to the forward Fourier transform of the nonkernel argument, rather than the inverse Fourier transform. This modification eliminates the need for the integration of the nonkernel, cross section–based function, which is a step that posed challenges for certain pointwise cross-section representations. It also removes the requirement for cross-section linearization. Optimized for graphics processing unit architectures, the approach significantly improves computational performance. These advancements are currently under evaluation as the foundation for the next-generation thermonuclear data file processing codes being developed at Lawrence Livermore National Laboratory.

Nuclear science and engineering↗

SU-E-T-536: Inhomogeneity Correction in Planning of Gamma Knife Treatments for Acoustic Schwannoma

Purpose: To find out the dose difference on targets and organs at risk for the treatment of acoustic schwannoma if the inhomogeneity correction (Convolution algorithm) is applied. Methods: Images of patients treated for acoustic schwannoma with Gamma Knife using TMR 10 algorithm were retrieved from database and replanned with Convolution and TMR 10 algorithm respectively. These patients were treated using a preplan scheme in following: (1) Before the actual treatment day, using the MRI image that was taken without a head frame on the patient's skull, a pre-treatment plan was made based on the default skull coordinates in the Gamma Knife treatment planning system (LGP); (2) then on treatment day, a head frame was placed on the patient's skull, and a CT image was taken. The CT image with head frame was registered and fused with the completed preplan; (3) the treatment plan was finalized and the treatment was delivered. To find out the dosimetry impact of inhomogeneity correction, we used the retrieved CT images to replan the treatment using Convolution algorithm in LGP software version 10.1.1. The dose distributions and the dose volume histograms for targets and OARs were compared for these two dose calculation algorithms. Results: The dose calculated with the Convolution algorithm in general is slightly lower than the one from TMR 10 around the boney area. The effect from the inhomogeneity correction is observable but not significant, and varies with the location of the tumor. Conclusion: Inhomogeneity correction slightly improve the dose accuracy for acoustic schwannoma Gamma Knife treatments although the correction may not be very significant. Our Result provides evidence for dose prescription adjustment to treat acoustic schwannoma. The actual clinical outcome of switching from using TMR10 to using Convolution needs to be further investigated.

60 APPLIED LIFE SCIENCES↗

SU-E-CAMPUS-I-06: Y90 PET/CT for the Instantaneous Determination of Both Target and Non-Target Absorbed Doses Following Hepatic Radioembolization

Purpose The process of converting Yttrium-90 (Y90) PET/CT images into 3D absorbed dose maps will be explained. The simple methods presented will allow the medical physicst to analyze Y90 PET images following radioembolization and determine the absorbed dose to tumor, normal liver parenchyma and other areas of interest, without application of Monte-Carlo radiation transport or dose-point-kernel (DPK) convolution. Methods Absorbed dose can be computed from Y90 PET/CT images based on the premise that radioembolization is a permanent implant with a constant relative activity distribution after infusion. Many Y90 PET/CT publications have used DPK convolution to obtain 3D absorbed dose maps. However, this method requires specialized software limiting clinical utility. The Local Deposition method, an alternative to DPK convolution, can be used to obtain absorbed dose and requires no additional computer processing. Pixel values from regions of interest drawn on Y90 PET/CT images can be converted to absorbed dose (Gy) by multiplication with a scalar constant. Results There is evidence that suggests the Local Deposition method may actually be more accurate than DPK convolution and it has been successfully used in a recent Y90 PET/CT publication. We have analytically compared dose-volume-histograms (DVH) for phantom hot-spheres to determine the difference between the DPK and Local Deposition methods, as a function of PET scanner point-spread-function for Y90. We have found that for PET/CT systems with a FWHM greater than 3.0 mm when imaging Y90, the Local Deposition Method provides a more accurate representation of DVH, regardless of target size than DPK convolution. Conclusion Using the Local Deposition Method, post-radioembolization Y90 PET/CT images can be transformed into 3D absorbed dose maps of the liver. An interventional radiologist or a Medical Physicist can perform this transformation in a clinical setting, allowing for rapid prediction of treatment efficacy by comparison to published tumoricidal thresholds.

60 APPLIED LIFE SCIENCES↗

Comparison of Shallow and Deep Learning Methods on Classifying the Regional Pattern of Diffuse Lung Disease

This study aimed to compare shallow and deep learning of classifying the patterns of interstitial lung diseases (ILDs). Using high-resolution computed tomography images, two experienced radiologists marked 1200 regions of interest (ROIs), in which 600 ROIs were each acquired using a GE or Siemens scanner and each group of 600 ROIs consisted of 100 ROIs for subregions that included normal and five regional pulmonary disease patterns (ground-glass opacity, consolidation, reticular opacity, emphysema, and honeycombing). We employed the convolution neural network (CNN) with six learnable layers that consisted of four convolution layers and two fully connected layers. The classification results were compared with the results classified by a shallow learning of a support vector machine (SVM). The CNN classifier showed significantly better performance for accuracy compared with that of the SVM classifier by 6–9%. As the convolution layer increases, the classification accuracy of the CNN showed better performance from 81.27 to 95.12%. Especially in the cases showing pathological ambiguity such as between normal and emphysema cases or between honeycombing and reticular opacity cases, the increment of the convolution layer greatly drops the misclassification rate between each case. Conclusively, the CNN classifier showed significantly greater accuracy than the SVM classifier, and the results implied structural characteristics that are inherent to the specific ILD patterns.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

A Bayesian Deep Learning Approach to Near-Term Climate Prediction

Since model bias and associated initialization shock are serious shortcomings that reduce prediction skills in state-of-the-art decadal climate prediction efforts, we pursue a complementary machine-learning-based approach to climate prediction. The example problem setting we consider consists of predicting natural variability of the North Atlantic sea surface temperature on the interannual timescale in the pre-industrial control simulation of the Community Earth System Model. While previous works have considered the use of recurrent networks such as convolutional LSTMs and reservoir computing networks in this and other similar problem settings, we currently focus on the use of feedforward convolutional networks. In particular, we find that a feedforward convolutional network with a Densenet architecture is able to outperform a convolutional LSTM in terms of predictive skill. Next, we go on to consider a probabilistic formulation of the same network based on Stein variational gradient descent and find that in addition to providing useful measures of predictive uncertainty, the probabilistic (Bayesian) version improves on its deterministic counterpart in terms of predictive skill. Finally, we characterize the reliability of the ensemble of machine learning models obtained in the probabilistic setting by using analysis tools developed in the context of ensemble numerical weather prediction.

54 ENVIRONMENTAL SCIENCES↗

Uncertainty in the visibility mask of a survey and its effects on the clustering of biased tracers

The forecasted accuracy of upcoming surveys of large-scale structure cannot be achieved without a proper quantification of the error induced by foreground removal (or other systematics like 0-point photometry offset). Because these errors are highly correlated on the sky, their influence is expected to be especially important at very large scales, at and beyond the first Baryonic Acoustic Oscillation (BAO). In this work we quantify how the uncertainty in the visibility mask of a survey, that gives the survey depth in a specific sky area, influences the measured power spectrum of a sample of tracers of the density field and its covariance matrix. We start from a very large set of 10,000 catalogs of dark matter (DM) halos in periodic cosmological boxes, produced with the PINOCCHIO approximate method. To make an analytic approach feasible, we assume luminosity-independent halo bias and an idealized geometry for the visibility mask, that is constant in square tiles of physical length l ; this should be interpreted as the projection, at the observation redshift, of the angular correlation scale of the foreground residuals. We find that the power spectrum of these biased tracers can be expressed as the sum of a cosmological term, a mask term and a term involving their convolution. The mask and convolution terms scale like P ∝ l {sup 2}σ {sub A} {sup 2}, where σ {sub A} {sup 2} is the variance of the uncertainty on the visibility mask. With l = 30−100 Mpc/ h and σ {sub A} = 5−20%, the mask term can be significant at k ∼ 0.01−0.1 h /Mpc, and the convolution term can amount to ∼ 1−10% of the total. The influence of mask uncertainty on power spectrum covariance is more complicated: the coupling of the convolution term with the other two gives rise to several mixed terms, that we quantify by difference using the mock catalogs. These are found to be of the same order of the mask covariance, and to introduce non-diagonal terms at large scales. As a consequence, the power spectrum covariance matrix cannot be expressed as the sum of a cosmological and of a mask term. More realistic settings (realistic foregrounds, luminosity-dependent bias) make the analytical approach not feasible, and the problem requires on the one hand usage of extended sets of mock catalogs, on the other hand detailed knowledge of the correlations among errors in the visibility masks. Our results lie down the theoretical bases to quantify the impact that uncertainties in the mask calibration have on the derivation of cosmological constraints from large spectroscopic surveys.

79 ASTRONOMY AND ASTROPHYSICS↗

Scaling deep learning workloads: NVIDIA DGX-1/Pascal and Intel Knights Landing

Deep Learning (DL) algorithms have become ubiquitous in data analytics. As a result, major computing vendors --- including NVIDIA, Intel, AMD, and IBM --- have architectural road-maps influenced by DL workloads. Furthermore, several vendors have recently advertised new computing products as accelerating large DL workloads. Unfortunately, it is difficult for data scientists to quantify the potential of these different products. This paper provides a performance and power analysis of important DL workloads on two major parallel architectures: NVIDIA DGX-1 (eight Pascal P100 GPUs interconnected with NVLink) and Intel Knights Landing (KNL) CPUs interconnected with Intel Omni-Path or Cray Aries. Our evaluation consists of a cross section of convolutional neural net workloads: CifarNet, AlexNet, GoogLeNet, and ResNet50 topologies using the Cifar10 and ImageNet datasets. The workloads are vendor-optimized for each architecture. Our analysis indicates that although GPUs provide the highest overall performance, the gap can close for some convolutional networks; and the KNL can be competitive in performance/watt. We find that NVLink facilitates scaling efficiency on GPUs. However, its importance is heavily dependent on neural network architecture. Furthermore, for weak-scaling --- sometimes encouraged by restricted GPU memory --- NVLink is less important.

Gawande, Nitin A.↗

A Modified Sequence-to-point HVAC Load Disaggregation Algorithm

This paper presents a modified sequence-to-point (S2P) algorithm for disaggregating the heat, ventilation, and air conditioning (HVAC) load from the total building electricity consumption. The original S2P model is convolutional neural network (CNN) based, which uses load profiles as inputs. We propose three modifications. First, the input convolution layer is changed from 1D to 2D so that normalized temperature profiles are also used inputs to the S2P model. Second, a drop-out layer is added to improve adaptability and generalizability so that the model trained in one area can be transferred to other geographical areas without labelled HVAC data. Third, a fine-tuning process is proposed for areas with a small amount of labelled HVAC data so that the pre-trained S2P model can be fine-tuned to achieve higher disaggregation accuracy (i.e., better transferability) in other areas. The model is first trained and tested using smart meter and sub-metered HVAC data collected in Austin, Texas. Then, the trained model is tested on two other areas: Boulder, Colorado and San Diego, California. Simulation results show that the proposed modified S2P algorithm outperforms the original S2P model and the support-vector machine based approach in accuracy, adaptability, and transferability.

Ye, Kai↗

A nonvoxel-based dose convolution/superposition algorithm optimized for scalable GPU architectures

Purpose: Real-time adaptive planning and treatment has been infeasible due in part to its high computational complexity. There have been many recent efforts to utilize graphics processing units (GPUs) to accelerate the computational performance and dose accuracy in radiation therapy. Data structure and memory access patterns are the key GPU factors that determine the computational performance and accuracy. In this paper, the authors present a nonvoxel-based (NVB) approach to maximize computational and memory access efficiency and throughput on the GPU. Methods: The proposed algorithm employs a ray-tracing mechanism to restructure the 3D data sets computed from the CT anatomy into a nonvoxel-based framework. In a process that takes only a few milliseconds of computing time, the algorithm restructured the data sets by ray-tracing through precalculated CT volumes to realign the coordinate system along the convolution direction, as defined by zenithal and azimuthal angles. During the ray-tracing step, the data were resampled according to radial sampling and parallel ray-spacing parameters making the algorithm independent of the original CT resolution. The nonvoxel-based algorithm presented in this paper also demonstrated a trade-off in computational performance and dose accuracy for different coordinate system configurations. In order to find the best balance between the computed speedup and the accuracy, the authors employed an exhaustive parameter search on all sampling parameters that defined the coordinate system configuration: zenithal, azimuthal, and radial sampling of the convolution algorithm, as well as the parallel ray spacing during ray tracing. The angular sampling parameters were varied between 4 and 48 discrete angles, while both radial sampling and parallel ray spacing were varied from 0.5 to 10 mm. The gamma distribution analysis method (γ) was used to compare the dose distributions using 2% and 2 mm dose difference and distance-to-agreement criteria, respectively. Accuracy was investigated using three distinct phantoms with varied geometries and heterogeneities and on a series of 14 segmented lung CT data sets. Performance gains were calculated using three 256 mm cube homogenous water phantoms, with isotropic voxel dimensions of 1, 2, and 4 mm. Results: The nonvoxel-based GPU algorithm was independent of the data size and provided significant computational gains over the CPU algorithm for large CT data sizes. The parameter search analysis also showed that the ray combination of 8 zenithal and 8 azimuthal angles along with 1 mm radial sampling and 2 mm parallel ray spacing maintained dose accuracy with greater than 99% of voxels passing the γ test. Combining the acceleration obtained from GPU parallelization with the sampling optimization, the authors achieved a total performance improvement factor of >175 000 when compared to our voxel-based ground truth CPU benchmark and a factor of 20 compared with a voxel-based GPU dose convolution method. Conclusions: The nonvoxel-based convolution method yielded substantial performance improvements over a generic GPU implementation, while maintaining accuracy as compared to a CPU computed ground truth dose distribution. Such an algorithm can be a key contribution toward developing tools for adaptive radiation therapy systems.

60 APPLIED LIFE SCIENCES↗

TU-D-209-02: A Backscatter Point Spread Function for Entrance Skin Dose Determination

Purpose: To determine the distribution of backscattered radiation to the skin resulting from a non-uniform distribution of primary radiation through convolution with a backscatter point spread function (PSF). Methods: A backscatter PSF is determined using Monte Carlo simulation of a 1 mm primary beam incident on a 30 × 30 cm × 20 cm thick PMMA phantom using EGSnrc software. A primary profile is similarly obtained without the phantom and the difference from the total provides the backscatter profile. This scatter PSF characterizes the backscatter spread for a “point” primary interaction and can be convolved with the entrance primary dose distribution to obtain the total entrance skin dose. The backscatter PSF was integrated into the skin dose tracking system (DTS), a graphical utility for displaying the color-coded skin dose distribution on a 3D graphic of the patient during interventional fluoroscopic procedures. The backscatter convolution method was validated for the non-uniform beam resulting from the use of an ROI attenuator. The ROI attenuator is a copper sheet with about 20% primary transmission (0.7 mm thick) containing a circular aperture; this attenuator is placed in the beam to reduce dose in the periphery while maintaining full dose in the region of interest. The DTS calculated primary plus backscatter distribution is compared to that measured with GafChromic film and that calculated using EGSnrc Monte-Carlo software. Results: The PSF convolution method used in the DTS software was able to account for the spread of backscatter from the ROI region to the region under the attenuator. The skin dose distribution determined using DTS with the ROI attenuator was in good agreement with the distributions measured with Gafchromic film and determined by Monte Carlo simulation Conclusion: The PSF convolution technique provides an accurate alternative for entrance skin dose determination with non-uniform primary x-ray beams. Partial support from NIH Grant R01-EB002873 and Toshiba Medical Systems Corp.

60 APPLIED LIFE SCIENCES↗