Search NASA⌕ Search

SEARCH · Search NASA

Results for “randomized algorithm”

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 1,621 records · Page 90

Active Learning with Irrelevant Examples

Active learning algorithms attempt to accelerate the learning process by requesting labels for the most informative items first. In real-world problems, however, there may exist unlabeled items that are irrelevant to the user's classification goals. Queries about these points slow down learning because they provide no information about the problem of interest. We have observed that when irrelevant items are present, active learning can perform worse than random selection, requiring more time (queries) to achieve the same level of accuracy. Therefore, we propose a novel approach, Relevance Bias, in which the active learner combines its default selection heuristic with the output of a simultaneously trained relevance classifier to favor items that are likely to be both informative and relevant. In our experiments on a real-world problem and two benchmark datasets, the Relevance Bias approach significantly improved the learning rate of three different active learning approaches.

machine learning↗

SU-C-BRB-03: Cross-Institutional Validation of An Ultrafast Automatic Planning Platform for Breast Irradiation

Purpose: Recently an ultrafast automatic planning system for breast irradiation using tangential beams was developed by modeling relationships between patient anatomy and achieved dose distribution. This study evaluates the performance of this system when applied to a different patient population and dose calculation algorithm. Methods: The system and its anatomy-to-dose models was developed at institution A based on 20 cases, which were planned using manual fluence painting technique and calculated WITH heterogeneity correction. Institution B uses field-in-field planning technique and dose calculation WITHOUT heterogeneity correction. 11 breast cases treated at Institution B were randomly selected for retrospective study, including left and right sides, and different breast size (irradiated volumes defined by Jaw/MLC opening range from 875cc to 3516cc). Comparisons between plans generated automatically (Auto-Plans) and those used for treatment (Clinical-Plans) included: energy choice (single/mixed), volumes receiving 95%/100%/105%/110% Rx dose (V95%/V100%/V105%/V100%) relative to irradiated volume, D1cc, and LungV20Gy. Results: In 9 out of 11 cases single/mixed energy choice made by the software agreed with Clinical-Plans. For the remaining 2 cases software recommended using mixed energy and dosimetric improvements were observed. V100% were similar (p=0.223, Wilcoxon Signed-Rank test) between Auto-Plans and Clinical-Plans (57.6±8.9% vs. 54.8±9.5%). V95% is 2.3±3.0% higher for Auto-Plans (p=0.027), indicating reduced cold areas. Hot spot volume V105% were significantly reduced in Auto-Plan by 14.4±7.2% (p=0.004). Absolute V105% was reduced from 395.6±359.9cc for Clinical-Plans to 108.7±163cc for Auto-Plans. D1cc was 107.4±2.8% for Auto-Plans, and 109.2±2.4% for Clinical-Plans (p=0.056). LungV20Gy were 13.6±4.0% for Auto-Plan vs. 14.0±4.1% for Clinical-Plans (p=0.043). All optimizations were finished within 1.5min. Conclusion: The performance of this breast auto-planning system remained stable and satisfactory when applied to a different patient population and dose calculation algorithm. The auto-planning system was able to produce clinically similar Rx dose coverage with significantly improved homogeneity inside breast tissue, in less than 1.5min.

60 APPLIED LIFE SCIENCES↗

Using machine learning to predict the correlation of spectra using SDSS magnitudes as an improvement to the Locus Algorithm

The Locus Algorithm is a new technique to improve the quality of differential photometry by optimising the choices of reference stars. At the heart of this algorithm is a routine to assess how good each potential reference star is by comparing its SDSS magnitude values to those of the target star. In this way, the difference in wavelength-dependent effects of the Earth’s atmospheric scattering between target and reference can be minimised. This paper sets out a new way to estimate the quality of each reference star using machine learning. A random subset of stars from SDSS with spectra was chosen. For each one, a suitable reference star, also with a spectrum, was chosen. The correlation between the two spectra in the SDSS r band (between 550 nm and 700 nm) was taken to be the gold-standard measure of how well they match up for differential photometry. The five SDSS magnitude values for each of these stars were used as predictors. A number of supervised machine learning models were constructed on a training set of the stars and were each evaluated on a testing set. The model using Support Vector Regression had the best performance of these models. It was then tested on a final, hold-out, validation set of stars to get an unbiased measure of its performance. With an R 2 of 0.62, the SVR model presents enhanced performance for the Locus Algorithm technique.

79 ASTRONOMY AND ASTROPHYSICS↗

SPRED: A machine learning approach for the identification of classical and non-classical secretory proteins in mammalian genomes

Eukaryotic protein secretion generally occurs via the classical secretory pathway that traverses the ER and Golgi apparatus. Secreted proteins usually contain a signal sequence with all the essential information required to target them for secretion. However, some proteins like fibroblast growth factors (FGF-1, FGF-2), interleukins (IL-1 alpha, IL-1 beta), galectins and thioredoxin are exported by an alternative pathway. This is known as leaderless or non-classical secretion and works without a signal sequence. Most computational methods for the identification of secretory proteins use the signal peptide as indicator and are therefore not able to identify substrates of non-classical secretion. In this work, we report a random forest method, SPRED, to identify secretory proteins from protein sequences irrespective of N-terminal signal peptides, thus allowing also correct classification of non-classical secretory proteins. Training was performed on a dataset containing 600 extracellular proteins and 600 cytoplasmic and/or nuclear proteins. The algorithm was tested on 180 extracellular proteins and 1380 cytoplasmic and/or nuclear proteins. We obtained 85.92% accuracy from training and 82.18% accuracy from testing. Since SPRED does not use N-terminal signals, it can detect non-classical secreted proteins by filtering those secreted proteins with an N-terminal signal by using SignalP. SPRED predicted 15 out of 19 experimentally verified non-classical secretory proteins. By scanning the entire human proteome we identified 566 protein sequences potentially undergoing non-classical secretion. The dataset and standalone version of the SPRED software is available at (http://www.inb.uni-luebeck.de/tools-demos/spred/spred).

60 APPLIED LIFE SCIENCES↗

A Comprehensive Comparison of IMRT and VMAT Plan Quality for Prostate Cancer Treatment

Purpose: We performed a comprehensive comparative study of the plan quality between volumetric-modulated arc therapy (VMAT) and intensity-modulated radiation therapy (IMRT) for the treatment of prostate cancer. Methods and Materials: Eleven patients with prostate cancer treated at our institution were randomly selected for this study. For each patient, a VMAT plan and a series of IMRT plans using an increasing number of beams (8, 12, 16, 20, and 24 beams) were examined. All plans were generated using our in-house-developed automatic inverse planning (AIP) algorithm. An existing eight-beam clinical IMRT plan, which was used to treat the patient, was used as the reference plan. For each patient, all AIP-generated plans were optimized to achieve the same level of planning target volume (PTV) coverage as the reference plan. Plan quality was evaluated by measuring mean dose to and dose-volume statistics of the organs at risk, especially the rectum, from each type of plan. Results: For the same PTV coverage, the AIP-generated VMAT plans had significantly better plan quality in terms of rectum sparing than the eight-beam clinical and AIP-generated IMRT plans (p < 0.0001). However, the differences between the IMRT and VMAT plans in all the dosimetric indices decreased as the number of beams used in IMRT increased. IMRT plan quality was similar or superior to that of VMAT when the number of beams in IMRT was increased to a certain number, which ranged from 12 to 24 for the set of patients studied. The superior VMAT plan quality resulted in approximately 30% more monitor units than the eight-beam IMRT plans, but the delivery time was still less than 3 min. Conclusions: Considering the superior plan quality as well as the delivery efficiency of VMAT compared with that of IMRT, VMAT may be the preferred modality for treating prostate cancer.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Polynomial chaos expansions on principal geodesic Grassmannian submanifolds for surrogate modeling and uncertainty quantification

In this work we introduce a manifold learning-based surrogate modeling framework for uncertainty quantification in high-dimensional stochastic systems. Our first goal is to perform data mining on the available simulation data to identify a set of low-dimensional (latent) descriptors that efficiently parameterize the response of the high-dimensional computational model. To this end, we employ Principal Geodesic Analysis on the Grassmann manifold of the response to identify a set of disjoint principal geodesic submanifolds, of possibly different dimension, that captures the variation in the data. Since operations on the Grassmann require the data to be concentrated, we propose an adaptive algorithm based on Riemannian K-means and the minimization of the sample Fréchet variance on the Grassmann manifold to identify “local” principal geodesic submanifolds that represent different system behavior across the parameter space. Polynomial chaos expansion is then used to construct a mapping between the random input parameters and the projection of the response on these local principal geodesic submanifolds. Here, the method is demonstrated on four test cases, a toy-example that involves points on a hypersphere, a Lotka-Volterra dynamical system, a continuous-flow stirred-tank chemical reactor system, and a two-dimensional Rayleigh-Bénard convection problem.

42 ENGINEERING↗

Power Grid Reliability Estimation via Adaptive Importance Sampling

Electricity production currently generates approximately 25% of greenhouse gas emissions in the USA. Thus, increasing the amount of renewable energy is a key step to carbon neutrality. However, integrating a large amount of fluctuating renewable generation is a significant challenge for power grid operating and planning. Grid reliability, i.e., an ability to meet operational constraints under power fluctuations, is probably the most important of them. In this letter, we propose computationally efficient and accurate methods to estimate the probability of line overflow, i.e., reliability constraints violation, under a known distribution of renewable energy generation. To this end, we investigate an importance sampling approach, a flexible extension of Monte-Carlo methods, which adaptively changes the sampling distribution to generate more samples near the reliability boundary. The approach allows to estimate overload probability in real-time based only on a few dozens of random samples, compared to thousands required by the plain Monte-Carlo. Our study focuses on high voltage direct current power transmission grids with linear reliability constraints on power injections and line currents. Herein, we propose a novel theoretically justified physics-informed adaptive importance sampling algorithm and compare its performance to state-of-the-art methods on multiple IEEE power grid test cases.

power system faults↗

Enhancing scalability of a matrix-free eigensolver for studying many-body localization

We propose several techniques to enhance the parallel scalability of a matrix-free eigensolver designed for studying many-body localization (MBL) of quantum spin chain models with nearest-neighbor interactions and on-site disorder. This type of problem is computationally challenging because the dimension of the associated Hamiltonian matrix grows exponentially with respect to the number of spins L, and we need to average over different realizations of the random disorder to obtain relevant statistical behavior. For each disorder realization, we need to compute eigenvalues from different regions of the spectrum and their corresponding eigenvectors. In previous work, the interior eigenstates for a single eigenvalue problem are computed via the shift-and-invert Lanczos algorithm. Due to the extremely high memory footprint of the LU factorizations, this technique is not well suited for large L’s. For example, we need thousands of compute nodes on modern high performance computing infrastructures to go beyond L = 24. The matrix-free approach does not suffer from this memory bottleneck, however, its scalability is limited by a computation and communication load imbalance. To reduce this imbalance and to significantly enhance the scalability of the matrix-free eigensolver, we reorder the matrix and leverage the consistent space runtime, CSPACER. We also show its efficiency in managing irregular communication patterns at scale compared to optimized MPI non-blocking two-sided and one-sided RMA implementation variants. This effort enables us to study MBL for spin chains with a larger number of spins. The efficiency and effectiveness of the proposed algorithm is demonstrated by computing eigenstates on a massively parallel many-core high performance computer.

METIS↗

A numerical homogenization method for heterogeneous, anisotropic elastic media based on multiscale theory

The development of reliable methods for upscaling fine scale models of elastic media has long been an important topic for rock physics and applied seismology. Several effective medium theories have been developed to provide elastic parameters for materials such as finely layered media or randomly oriented or aligned fractures. In such cases, the analytic solutions for upscaled properties can be used for accurate prediction of wave propagation. However, such theories cannot be applied directly to homogenize elastic media with more complex, arbitrary spatial heterogeneity. We therefore propose a numerical homogenization algorithm based on multiscale finite element methods for simulating elastic wave propagation in heterogeneous, anisotropic elastic media. Specifically, our method used multiscale basis functions obtained from a local linear elasticity problem with appropriately defined boundary conditions. Homogenized, effective medium parameters were then computed using these basis functions, and the approach applied a numerical discretization that is similar to the rotated staggered-grid finite difference scheme. Comparisons of the results from our method and from conventional, analytical approaches for finely layered media showed that the homogenization reliably estimated elastic parameters for this simple geometry. Additional tests examined anisotropic models with arbitrary spatial heterogeneity where the average size of the heterogeneities ranged from several centimeters to several meters, and the ratio between the dominant wavelength and the average size of the arbitrary heterogeneities ranged from 10 to 100. Comparisons to finite-difference simulations proved that the numerical homogenization was equally accurate for these complex cases.

58 GEOSCIENCES↗

A procedure used for a ground truth study of a land use map of North Alabama generated from LANDSAT data

A land use map of a five county area in North Alabama was generated from LANDSAT data using a supervised classification algorithm. There was good overall agreement between the land use designated and known conditions, but there were also obvious discrepancies. In ground checking the map, two types of errors were encountered - shift and misclassification - and a method was developed to eliminate or greatly reduce the errors. Randomly selected study areas containing 2,525 pixels were analyzed. Overall, 76.3 percent of the pixels were correctly classified. A contingency coefficient of correlation was calculated to be 0.7 which is significant at the alpha = 0.01 level. The land use maps generated by computers from LANDSAT data are useful for overall land use by regional agencies. However, care must be used when making detailed analysis of small areas. The procedure used for conducting the ground truth study together with data from representative study areas is presented.

Downs, S. W., Jr.↗

Structured codebook design in CELP

Codebook Excited Linear Protection (CELP) is a popular analysis by synthesis technique for quantizing speech at bit rates from 4 to 6 kbps. Codebook design techniques to date have been largely based on either random (often Gaussian) codebooks, or on known binary or ternary codes which efficiently map the space of (assumed white) excitation codevectors. It has been shown that by introducing symmetries into the codebook, good complexity reduction can be realized with only marginal decrease in performance. Codebook design algorithms are considered for a wide range of structured codebooks.

Leblanc, W. P.↗

Band-Moment Compression of AVIRIS Hyperspectral Data and its Use in the Detection of Vegetation Stress

A remote sensing campaign was conducted over a U.S. Department of Agriculture test farm at Shelton, Nebraska. An experimental field was set off in plots that were differentially treated with anhydrous ammonia. Four replicates of 0-kg/ha to 200-kg/ha plots, in 50-kg/ha increments, were set out in a random block design. Low-altitude (GSD of 3 m) Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) hyperspectral data were collected over the site in 224 bands. Simultaneously, ground data were collected to support the airborne imagery. In an effort to reduce data load while maintaining or enhancing algorithm performance for vegetation stress detection, band-moment compression and analysis was applied to the AVIRIS image cube. The results indicated that band-moment techniques compress the AVIRIS dataset significantly while retaining the capability of detecting environmentally induced vegetation stress.

Estep, L.↗

Linear Subpixel Learning Algorithm for Land Cover Classification from WELD using High Performance Computing

In this work, we use a Fully Constrained Least Squares Subpixel Learning Algorithm to unmix global WELD (Web Enabled Landsat Data) to obtain fractions or abundances of substrate (S), vegetation (V) and dark objects (D) classes. Because of the sheer nature of data and compute needs, we leveraged the NASA Earth Exchange (NEX) high performance computing architecture to optimize and scale our algorithm for large-scale processing. Subsequently, the S-V-D abundance maps were characterized into 4 classes namely, forest, farmland, water and urban areas (with NPP-VIIRS-national polar orbiting partnership visible infrared imaging radiometer suite nighttime lights data) over California, USA using Random Forest classifier. Validation of these land cover maps with NLCD (National Land Cover Database) 2011 products and NAFD (North American Forest Dynamics) static forest cover maps showed that an overall classification accuracy of over 91 percent was achieved, which is a 6 percent improvement in unmixing based classification relative to per-pixel-based classification. As such, abundance maps continue to offer an useful alternative to high-spatial resolution data derived classification maps for forest inventory analysis, multi-class mapping for eco-climatic models and applications, fast multi-temporal trend analysis and for societal and policy-relevant applications needed at the watershed scale.

Subpixel↗

Automatic variable selection in ecological niche modeling: A case study using Cassin’s Sparrow (Peucaea cassinii)

MERRA/Max provides a feature selection approach to dimensionality reduction that enables direct use of global climate model outputs in ecological niche modeling. The system accomplishes this reduction through a Monte Carlo optimization in which many independent MaxEnt runs, operating on a species occurrence file and a small set of randomly selected variables in a large collection of variables, converge on an estimate of the top contributing predictors in the larger collection. These top predictors can be viewed as potential candidates in the variable selection step of the ecological niche modeling process. MERRA/Max’s Monte Carlo algorithm operates on files stored in the underlying filesystem, making it scalable to large data sets. Its software components can run as parallel processes in a high-performance cloud computing environment to yield near real-time performance. In tests using Cassin’s Sparrow (Peucaea cassinii) as the target species, MERRA/Max selected a set of predictors from Worldclim’s Bioclim collection of 19 environmental variables that have been shown to be important determinants of the species’ bioclimatic niche. It also selected biologically and ecologically plausible predictors from a more diverse set of 86 environmental variables derived from NASA’s Modern-Era Retrospective Analysis for Research and Applications Version 2 (MERRA-2) reanalysis, an output product of the Goddard Earth Observing System Version 5 (GEOS-5) modeling system. We believe these results point to a technological approach that could expand the use global climate model outputs in ecological niche modeling, foster exploratory experimentation with otherwise difficult-to-use climate data sets, streamline the modeling process, and, eventually, enable automated bioclimatic modeling as a practical, readily accessible, low-cost, commercial cloud service.

John L. Schnase↗

Implementation of Combinatorial Optimization Techniques for Automated Fiber Placement Through Thickness Defect Stack-Up Minimization

The Computer Aided Process Planning (CAPP) module was developed to facilitate and accelerate the process planning workflow for Automated Fiber Placement (AFP). CAPP assists process planners in identifying optimal starting point locations and layup strategies for each ply of a laminate. Ply optimization operates on measurement and scoring of geometry-based defects such as gaps, overlaps, angle deviation, and steering. This paper expands on the established framework for analyzing defect stack-up through thickness of a laminate. Four different combinatorial optimization algorithms are implemented and evaluated: (1) genetic algorithm, (2) differential evolution, (3) particle swarm, and (4) greedy search. The algorithms identify the optimal combination of ply-level layup strategies, by scoring potential laminates on defect stacking, using two different objective functions. A final optimization approach is also presented which trades some performance for a large gain in efficiency. These approaches are compared to a randomized combination using a complex tool surface in a virtual case study. The result is a streamlined methodology for comparing different laminate-level manufacturing strategies and minimizing the through thickness defect stack up.

CAPP↗

Single-Shot Femtosecond X-ray Diffraction from Randomly Oriented Ellipsoidal Nanoparticles

Coherent diffractive imaging of single particles using the single-shot 'diffract and destroy' approach with an x-ray free electron laser (FEL) was recently demonstrated. A high-resolution low-noise coherent diffraction pattern, representative of the object before it turns into a plasma and explodes, results from the interaction of the FEL with the particle. Iterative phase retrieval algorithms are used to reconstruct two-dimensional projection images of the object from the recorded intensities alone. Here we describe the first single-shot diffraction data set that mimics the data proposed for obtaining 3D structure from identical particles. Ellipsoidal iron oxide nanoparticles (250 nm x 50 nm) were aerosolized and injected through an aerodynamic lens stack into a soft x-ray FEL. Particle orientation was not controlled with this injection method. We observed that, at the instant the x-ray pulse interacts with the particle, a snapshot of the particle's orientation is encoded in the diffraction pattern. The results give credence to one of the technical concepts of imaging individual nanometer and subnanometer-sized objects such as single molecules or larger clusters of molecules using hard x-ray FELs and will be used to help develop robust algorithms for determining particle orientations and 3D structure.

43 PARTICLE ACCELERATORS↗

SU-F-T-609: Impact of Dosimetric Variation for Prescription Dose Using Analytical Anisotropic Algorithm (AAA) in Lung SBRT

Purpose: Actual irradiated prescription dose to patients cannot be verified. Thus, independent dose verification and second treatment planning system are used as the secondary check. AAA dose calculation engine has contributed to lung SBRT. We conducted a multi-institutional study to assess variation of prescription dose for lung SBRT when using AAA in reference to using Acuros XB and Clarkson algorithm. Methods: Six institutes in Japan participated in this study. All SBRT treatments were planed using AAA in Eclipse and Adaptive Convolve (AC) in Pinnacle3. All of the institutes used a same independent dose verification software program (Simple MU Analysis: SMU, Triangle Product, Ishikawa, Japan), which implemented a Clarkson-based dose calculation algorithm using CT image dataset. A retrospective analysis for lung SBRT plans (73 patients) was performed to compute the confidence limit (CL, Average±2SD) in dose between the AAA and the SMU. In one of the institutes, a additional analysis was conducted to evaluate the variations between the AAA and the Acuros XB (AXB). Results: The CL for SMU shows larger systematic and random errors of 8.7±9.9 % for AAA than the errors of 5.7±4.2 % for AC. The variations of AAA correlated with the mean CT values in the voxels of PTV (a correlation coefficient : −0.7) . The comparison of AXB vs. AAA shows smaller systematic and random errors of −0.7±1.7%. The correlation between dose variations for AXB and the mean CT values in PTV was weak (0.4). However, there were several plans with more than 2% deviation of AAPM TG114 (Maximum: −3.3 %). Conclusion: In comparison for AC, prescription dose calculated by AAA may be more variable in lung SBRT patient. Even AXB comparison shows unexpected variation. Care should be taken for the use of AAA in lung SBRT. This research is partially supported by Japan Agency for Medical Research and Development (AMED)

60 APPLIED LIFE SCIENCES↗

Predicting the fidelity of JPEG2000 compressed CT images using DICOM header information

Purpose: To propose multiple logistic regression (MLR) and artificial neural network (ANN) models constructed using digital imaging and communications in medicine (DICOM) header information in predicting the fidelity of Joint Photographic Experts Group (JPEG) 2000 compressed abdomen computed tomography (CT) images. Methods: Our institutional review board approved this study and waived informed patient consent. Using a JPEG2000 algorithm, 360 abdomen CT images were compressed reversibly (n = 48, as negative control) or irreversibly (n = 312) to one of different compression ratios (CRs) ranging from 4:1 to 10:1. Five radiologists independently determined whether the original and compressed images were distinguishable or indistinguishable. The 312 irreversibly compressed images were divided randomly into training (n = 156) and testing (n = 156) sets. The MLR and ANN models were constructed regarding the DICOM header information as independent variables and the pooled radiologists' responses as dependent variable. As independent variables, we selected the CR (DICOM tag number: 0028, 2112), effective tube current-time product (0018, 9332), section thickness (0018, 0050), and field of view (0018, 0090) among the DICOM tags. Using the training set, an optimal subset of independent variables was determined by backward stepwise selection in a four-fold cross-validation scheme. The MLR and ANN models were constructed with the determined independent variables using the training set. The models were then evaluated on the testing set by using receiver-operating-characteristic (ROC) analysis regarding the radiologists' pooled responses as the reference standard and by measuring Spearman rank correlation between the model prediction and the number of radiologists who rated the two images as distinguishable. Results: The CR and section thickness were determined as the optimal independent variables. The areas under the ROC curve for the MLR and ANN predictions were 0.91 (95% CI; 0.86, 0.95) and 0.92 (0.87, 0.96), respectively. The correlation coefficients of the MLR and ANN predictions with the number of radiologists who responded as distinguishable were 0.76 (0.69, 0.82, p < 0.001) and 0.78 (0.71, 0.83, p < 0.001), respectively. Conclusions: The MLR and ANN models constructed using the DICOM header information offer promise in predicting the fidelity of JPEG2000 compressed abdomen CT images.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗