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 829 records · Page 46

THE GINI COEFFICIENT AS A MORPHOLOGICAL MEASUREMENT OF STRONGLY LENSED GALAXIES IN THE IMAGE PLANE

Characterization of the morphology of strongly lensed galaxies is challenging because images of such galaxies are typically highly distorted. Lens modeling and source plane reconstruction is one approach that can provide reasonably undistorted images from which morphological measurements can be made, though at the expense of a highly spatially variable telescope point-spread function (PSF) when mapped back to the source plane. Unfortunately, modeling the lensing mass is a time- and resource-intensive process, and in many cases there are too few constraints to precisely model the lensing mass. If, however, useful morphological measurements could be made in the image plane rather than the source plane, it would bypass this issue and obviate the need for a source reconstruction process for some applications. We examine the use of the Gini coefficient as one such measurement. Because it depends on the cumulative distribution of the light of a galaxy, but not the relative spatial positions, the fact that surface brightness is conserved by lensing means that the Gini coefficient may be well preserved by strong gravitational lensing. Through simulations, we test the extent to which the Gini coefficient is conserved, including by effects due to PSF convolution and pixelization, to determine whether it is invariant enough under lensing to be used as a measurement of galaxy morphology that can be made in the image plane.

79 ASTRONOMY AND ASTROPHYSICS↗

Photometric Classification of Early-time Supernova Light Curves with SCONE

Abstract In this work, we present classification results on early supernova light curves from SCONE, a photometric classifier that uses convolutional neural networks to categorize supernovae (SNe) by type using light-curve data. SCONE is able to identify SN types from light curves at any stage, from the night of initial alert to the end of their lifetimes. Simulated LSST SNe light curves were truncated at 0, 5, 15, 25, and 50 days after the trigger date and used to train Gaussian processes in wavelength and time space to produce wavelength–time heatmaps. SCONE uses these heatmaps to perform six-way classification between SN types Ia, II, Ibc, Ia-91bg, Iax, and SLSN-I. SCONE is able to perform classification with or without redshift, but we show that incorporating redshift information improves performance at each epoch. SCONE achieved 75% overall accuracy at the date of trigger (60% without redshift), and 89% accuracy 50 days after trigger (82% without redshift). SCONE was also tested on bright subsets of SNe ( r < 20 mag) and produced 91% accuracy at the date of trigger (83% without redshift) and 95% five days after trigger (94.7% without redshift). SCONE is the first application of convolutional neural networks to the early-time photometric transient classification problem. All of the data processing and model code developed for this paper can be found in the SCONE software package 1 1 github.com/helenqu/scone located at github.com/helenqu/scone (Qu 2021).

79 ASTRONOMY AND ASTROPHYSICS↗

X-Ray Spectro-polarimetry with Photoelectric Polarimeters

We derive a generalization of forward fitting for X-ray spectroscopy to include linear polarization of X-ray sources, appropriate for the anticipated next generation of space-based photoelectric polarimeters. We show that the inclusion of polarization sensitivity requires joint fitting to three observed spectra, one for each of the Stokes parameters, I(E), U(E), and Q(E). The equations for Stokes’ I(E) (the total intensity spectrum) are identical to the familiar case with no polarization sensitivity, and for which the model-predicted spectrum is obtained by a convolution of the source spectrum, F(E′), with the familiar energy response function, ϵ(E′) R(E′, E), where ϵ(E′) and R(E′, E) are the effective area and energy redistribution matrix, respectively. In addition to the energy spectrum, the two new relations for U(E) and Q(E) include the source polarization fraction and position angle versus energy, a(E′), and ψ{sub 0}{sup ′}(E{sup ′}), respectively, and the model-predicted spectra for these relations are obtained by a convolution with the “modulated” energy response function, μ(E′) ϵ (E′)R(E, E′), where μ(E′) is the energy-dependent modulation fraction that quantifies a polarimeter’s angular response to 100% polarized radiation. We present results of simulations with response parameters appropriate for the proposed PRAXyS Small Explorer observatory to illustrate the procedures and methods, and we discuss some aspects of photoelectric polarimeters with relevance to understanding their calibration and operation.

79 ASTRONOMY AND ASTROPHYSICS↗

Development of a Machine-Learned Cruise Guide Indicator for Rotorcraft

This paper presents a machine-learned virtual cruise guide indicator (vCGI) for Chinook helicopters. Two temporal neural networks were trained and evaluated on measured data from 55 flight tests, one for the fore rotor and another for the aft rotor, to predict a vCGI value, which protects 23 components from fatigue damage during steady-state conditions. Three different classes of machine learning architectures were evaluated for prediction of the vCGI from time sequences: a temporal convolutional neural network with 1D dilated causal convolutions, a long short-term memory recurrent neural network, and an attention-based transformer architecture. The final average model accuracy on unseen flight data is currently greater than 93% for CGI values which could result in fatigue damage and 90% for normal operation CGI values. Model accuracy was improved through a series of advancements in:(1) selection of optimal training data using temporal collective variables and unsupervised learning, (2) dataset augmentation with maximum-entropy temporal collective variables, and (3) implementation of a mixture-of-experts classification- regression approach using an adversarial classification approach to assign maneuver labels. The results are presented for each advancement in model development along with lessons learned in training machine learning models on real- world, time-dependent rotorcraft data.

Boyer, Mathew↗

DMTN-021: Implementation of Image Difference Decorrelation

Herein, we describe a method for decorrelating image differences produced by the Alard and Lupton (1998) method of PSF matching. Inspired by the recent work of Zackay et al. (2016) and the prior work of Kaiser [2004], this proposed method uses a single post-subtraction convolution of an image difference to remove the neighboring pixel covariances in the image difference that result from the convolution of the template image by the PSF matching kernel. We describe the method in detail, analyze its effects on image differences (both real and simulated) as well as on detections and photometry of detected sources in decorrelated image differences. We also compare the decorrelated image differences with those resulting from a basic implementation of Zackay et al. (2016). We describe the implementation of the new correction in the LSST image differencing pipeline, and discuss potential issues and areas of future research.

79 ASTRONOMY AND ASTROPHYSICS↗

Efficient algorithm for generating spectra using line-by-lne methods

A method is presented for efficient generation of spectra using line-by-line approaches. The only approximation is replacing the line shape function with an interpolation procedure, which makes the method independent of the line profile functional form. The resulting computational savings for large number of lines is proportional to the number of frequency points in the spectral range. Therefore, for large-scale problems the method can provide speedups of two orders of magnitude or more. A method was presented to generate line-by-line spectra efficiently. The first step was to replace the explicit calculation of the profile by the Newton divided-differences interpolating polynomial. The second step is to accumulate the lines effectively reducing their number to the number of frequency points. The final step is recognizing the resulting expression as a convolution and amenable to FFT methods. The reduction in computational effort for a configuration-to-configuration transition array with large number of lines is proportional to the number of frequency points. The method involves no approximations except for replacing the explicit profile evaluation by interpolation. Specifically, the line accumulation and convolution are exact given the interpolation procedure. Furthermore, the interpolation makes the method independent of the line profile functional form contrary to other schemes using FFT methods to generate line-by-line spectra but relying on the analytic form of the profile Fourier transform. Finally, the method relies on a uniform frequency mesh. For non-uniform frequency meshes, however, the method can be applied by using a suitable temporary uniform mesh and the results interpolated onto the final mesh with little additional cost.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Spectral Fluctuations of Incoherent Radiation And Measurement of Longitudinal Bunch Profile

A method for measurement of ultrashort beam current profile I{sub b}(t) is proposed that is based on detecting fluctuations of the spectral intensity P ({omega}) of single bunch incoherent radiation. We show that the variance of the Fourier transform of the spectrum is proportional to the convolution function of the beam current. After the convolution function is found, using phase retrieval technique one can restore the shape of the pulse in many practical cases.

43 PARTICLE ACCELERATORS↗

Variable-pulse-shape pulsed-power accelerator

A variable-pulse-shape pulsed-power accelerator is driven by a large number of independent LC drive circuits. Each LC circuit drives one or more coaxial transmission lines that deliver the circuit's output power to several water-insulated radial transmission lines that are connected in parallel at small radius by a water-insulated post-hole convolute. The accelerator can be impedance matched throughout. The coaxial transmission lines are sufficiently long to transit-time isolate the LC drive circuits from the water-insulated transmission lines, which allows each LC drive circuit to be operated without being affected by the other circuits. This enables the creation of any power pulse that can be mathematically described as a time-shifted linear combination of the pulses of the individual LC drive circuits. Therefore, the output power of the convolute can provide a variable pulse shape to a load that can be used for magnetically driven, quasi-isentropic compression experiments and other applications.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Artificial neural network based isotopic analysis of airborne radioactivity measurement for radiological incident detection

Responders need tools to rapidly detect and identify airborne alpha radioactivity during consequence management scenarios. Traditional continuous air monitor systems used for this purpose compute the net counts in various energy windows to determine the presence of specified isotopes, such as 235U, 239Pu, and 241Am. These calculations rely on having a well-calibrated detector, which is challenging in low-background environments. Here an alternative approach of using artificial neural networks to classify alpha spectra is presented. Two network architectures, fully connected and convolutional networks, were trained to classify alpha spectra into four categories: background and background plus the three isotopes above. Sources were injected into measured background at various fractions of the derived response level (DRL) corresponding to early-phase Protective Action Guides. The convolutional network identifies all sources at 1% of the DRL with average probability of detection of 95% and false alarm probability of 1%. Further, the network identifies sources ranging between 0.25% and 1% of the DRL with higher than 80% probability of detection and lower than 7% false alarm probability. Most significantly, the network performance improves in low-count background conditions, increasing its minimum probability of detection to 93% and reducing the false alarm probabilities to lower than 0.25%. These results show that, once trained on datasets representing a range of detection scenarios, artificial neural networks can accurately identify alpha isotopes of interest without the need for detector calibration.

Woldegiorgis, Surafel F.↗

Multi-View, Deep Learning, and Contextual Analysis: Promising Approaches for sUAS Land Cover Classification

Small Unmanned Aircraft System (sUAS) image collection and preprocessing to produce orthoimages, point clouds, and digital surface models have made great leaps forward in the past years. These advances and the wide spectrum of existing and emerging sUAS applications mandate parallel improvements in the algorithms used for image classification that take into consideration the specific characteristics of sUAS imagery. This chapter presents several alternative approaches to improve classification accuracy using redundant information in overlapped sUAS imagery captured in a typical sUAS mission. These include Deep Convolution Neural Network (DCNN) classification, Fully Convolutional Neural (FCN) Network, and contextual relationships under an Object-Based Image Analysis (OBIA) framework. The chapter synthesizes the results of implementing the introduced classification approaches using a case study of high-resolution sUAS images captured in a wetland area in Central Florida.

Liu, Tao↗

Device and method for constructing and displaying high quality images from imaging data by transforming a data structure utilizing machine learning techniques

Constructing a computer image from raw imaging data or encoded imaging data by transforming a first data structure in which the raw imaging data or the encoded imaging data is stored into a second data structure storing reorganized imaging data. The raw imaging data or the encoded imaging data is received, stored in the first data structure. The computer reorganizes the raw imaging data or the encoded imaging data into the reorganized data and stores the reorganized data in the second data structure, which is a multi-dimensional array having subarrays containing local information needed by a convolutional neural network for processing the reorganized data. Other portions of the multi-dimensional array store other portions of the raw imaging data or the encoded imaging data. The computer also processes the reorganized data using the convolutional neural network to construct the image, whereby a constructed image is formed.

Korbin, John P.↗

Graph Partitioning and Sparse Matrix Ordering using Reinforcement Learning and Graph Neural Networks

We present a novel method for graph partitioning, based on reinforcement learning and graph convolutional neural networks. Our approach is to recursively partition coarser representations of a given graph. The neural network is implemented using SAGE graph convolution layers, and trained using an advantage actor critic (A2C) agent. We present two variants, one for finding an edge separator that minimizes the normalized cut or quotient cut, and one that finds a small vertex separator. The vertex separators are then used to construct a nested dissection ordering to permute a sparse matrix so that its triangular factorization will incur less fill-in. The partitioning quality is compared with partitions obtained using METIS and SCOTCH, and the nested dissection ordering is evaluated in the sparse solver SuperLU. Our results show that the proposed method achieves similar partitioning quality as METIS and SCOTCH. Furthermore, the method generalizes across different classes of graphs, and works well on a variety of graphs from the SuiteSparse sparse matrix collection.

97 MATHEMATICS AND COMPUTING↗

Cosmic ray muon clustering for the MicroBooNE liquid argon time projection chamber using sMask-RCNN

In this article, we describe a modified implementation of Mask Region-based Convolutional Neural Networks (Mask-RCNN) for cosmic ray muon clustering in a liquid argon TPC and applied to MicroBooNE neutrino data. Our implementation of this network, called sMask-RCNN, uses sparse submanifold convolutions to increase processing speed on sparse datasets, and is compared to the original dense version in several metrics. The networks are trained to use wire readout images from the MicroBooNE liquid argon time projection chamber as input and produce individually labeled particle interactions within the image. These outputs are identified as either cosmic ray muon or electron neutrino interactions. We find that sMask-RCNN has an average pixel clustering efficiency of 85.9% compared to the dense network's average pixel clustering efficiency of 89.1%. We demonstrate the ability of sMask-RCNN used in conjunction with MicroBooNE's state-of-the-art Wire-Cell cosmic tagger to veto events containing only cosmic ray muons. The addition of sMask-RCNN to the Wire-Cell cosmic tagger removes 70% of the remaining cosmic ray muon background events at the same electron neutrino event signal efficiency. This event veto can provide 99.7% rejection of cosmic ray-only background events while maintaining an electron neutrino event-level signal efficiency of 80.1%. In addition to cosmic ray muon identification, sMask-RCNN could be used to extract features and identify different particle interaction types in other 3D-tracking detectors.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

WavPool: A New Block for Deep Neural Networks

Modern deep neural networks comprise many operational layers, such as dense or convolutional layers, which are often collected into blocks. In this work, we introduce a new, wavelet-transform-based network architecture that we call the multi-resolution perceptron: by adding a pooling layer, we create a new network block, the WavPool. The first step of the multi-resolution perceptron is transforming the data into its multi-resolution decomposition form by convolving the input data with filters of fixed coefficients but increasing size. Following image processing techniques, we are able to make scale and spatial information simultaneously accessible to the network without increasing the size of the data vector. WavPool outperforms a similar multilayer perceptron while using fewer parameters, and outperforms a comparable convolutional neural network by ~ 10% on relative accuracy on CIFAR-10.

McDermott, Samuel D.↗

Phosphor thermometry device for synchronized acquisition of luminescence lifetime decay and intensity on thermal barrier coatings

A phosphor thermometry device includes a laser that generates a laser pulse onto a thermal barrier coating (TBC) applied onto a substrate. A metallic bond coat layer is on the substrate. A ceramic top coat layer is on the bond coat layer and includes an undoped layer and a doped sensing layer having co-doped first and second rare-earth luminescent dopants that emit respective first and second different emission wavelengths upon excitation by the laser pulse. A detector receives reflected, convoluted luminescence signals from the TBC. First and second photomultiplier devices detect respective first and second different emission wavelengths of the convoluted luminescence signals. A controller receives and processes signals generated from respective first and second photomultiplier devices and determines luminescence lifetime decay and intensity variations for each of the respective first and second rare-earth luminescent dopants for temperature monitoring of the TBC.

Fouliard, Quentin↗

Deep learning model to detect various synchrophasor data anomalies

High-density synchrophasors provide valuable information for power grid situational awareness, operation and control. Unfortunately, due to factors including communication instability and hardware failure, their data quality can be greatly deteriorated by anomalies. Since the anomalies can impact the performance of the synchrophasor applications, it is of paramount significance to propose a model to detect anomalies in synchrophasor. In this study, a convolutional neural network model is established to detect and classify the anomalies in the synchrophasor measurements. Additionally, four types of anomalies observed in actual synchrophasors including erroneous patterns, random spikes, missing points and high-frequency interferences are considered in this study. The proposed model is extensively evaluated via field-collected measurements from the synchrophasor network in Jiangsu grid, China. The superior performance of the proposed model indicates the great potential of using deep learning for the detection of abnormal synchrophasor measurements.

42 ENGINEERING↗

Impurity gas detection for SNF canisters using probabilistic deep learning and acoustic sensing *

Abstract Monitoring impurity gases in spent nuclear fuel (SNF) canisters is a novel structural health monitoring approach for SNF in dry storage. The SNF canisters are sealed containers that do not facilitate visual access to the inside. Acoustic sensing can be deployed by taking advantage of the pathways unobstructed by internal hardware. Although the ultrasonic time-of-flight measurement can provide valuable information, it is limited in its ability to discern the concentration of only one impurity gas. As such, deep learning algorithms, particularly convolutional neural networks (CNNs), offer a promising solution. In this study, CNN-based probabilistic deep learning models were implemented to detect and quantify multiple impurity gases in helium. An experimental platform was established to simulate canister conditions, and ultrasonic test data were collected. The presence of argon and air in helium at concentrations ranging from 0% to 1.2% at increments of 0.05% was considered. The multi-layer perceptron, decision tree, and logistic regression classifiers achieved high accuracies when distinguishing pure helium from helium with impurities. CNN with dropout layers and CNN using maximum likelihood estimation showed a similar performance, indicating their ability to capture uncertainties. The ensemble CNN model exhibited improved predictions and the ability to balance individual gas concentration by integrating 1D- and 2D-CNN models. These findings contribute probabilistic deep learning solutions for impurity gas detection and analysis within SNF canisters, thus ensuring safe storage and management of SNFs.

47 OTHER INSTRUMENTATION↗

Scaling the training of particle classification on simulated MicroBooNE events to multiple GPUs

Measurements in Liquid Argon Time Projection Chamber (LArTPC) neutrino detectors, such as the MicroBooNE detector at Fermilab, feature large, high fidelity event images. Deep learning techniques have been extremely successful in classification tasks of photographs, but their application to LArTPC event images is challenging, due to the large size of the events. Events in these detectors are typically two orders of magnitude larger than images found in classical challenges, like recognition of handwritten digits contained in the MNIST database or object recognition in the ImageNet database. Ideally, training would occur on many instances of the entire event data, instead of many instances of cropped regions of interest from the event data. However, such efforts lead to extremely long training cycles, which slow down the exploration of new network architectures and hyperparameter scans to improve the classification performance. We present studies of scaling a LArTPC classification problem on multiple architectures, spanning multiple nodes. The studies are carried out on simulated events in the MicroBooNE detector. We emphasize that it is beyond the scope of this study to optimize networks or extract the physics from any results here. Institutional computing at Pacific Northwest National Laboratory and the SummitDev machine at Oak Ridge National Laboratory’s Leadership Computing Facility have been used. To our knowledge, this is the first use of state-of-the-art Convolutional Neural Networks for particle physics and their attendant compute techniques onto the DOE Leadership Class Facilities. We expect benefits to accrue particularly to the Deep Underground Neutrino Experiment (DUNE) LArTPC program, the flagship US High Energy Physics (HEP) program for the coming decades.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗