Search NASA⌕ Search

SEARCH · Search NASA

Results for “NEURAL NETWORK”

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 361 records · Page 20

Reconfigurable Control with Neural Network Augmentation for a Modified F-15 Aircraft

Description of the performance of a simplified dynamic inversion controller with neural network augmentation follows. Simulation studies focus on the results with and without neural network adaptation through the use of an F-15 aircraft simulator that has been modified to include canards. Simulated control law performance with a surface failure, in addition to an aerodynamic failure, is presented. The aircraft, with adaptation, attempts to minimize the inertial cross-coupling effect of the failure (a control derivative anomaly associated with a jammed control surface). The dynamic inversion controller calculates necessary surface commands to achieve desired rates. The dynamic inversion controller uses approximate short period and roll axis dynamics. The yaw axis controller is a sideslip rate command system. Methods are described to reduce the cross-coupling effect and maintain adequate tracking errors for control surface failures. The aerodynamic failure destabilizes the pitching moment due to angle of attack. The results show that control of the aircraft with the neural networks is easier (more damped) than without the neural networks. Simulation results show neural network augmentation of the controller improves performance with aerodynamic and control surface failures in terms of tracking error and cross-coupling reduction.

Burken, John J.↗

Adaptive Control Using Neural Network Augmentation for a Modified F-15 Aircraft

Description of the performance of a simplified dynamic inversion controller with neural network augmentation follows. Simulation studies focus on the results with and without neural network adaptation through the use of an F-15 aircraft simulator that has been modified to include canards. Simulated control law performance with a surface failure, in addition to an aerodynamic failure, is presented. The aircraft, with adaptation, attempts to minimize the inertial cross-coupling effect of the failure (a control derivative anomaly associated with a jammed control surface). The dynamic inversion controller calculates necessary surface commands to achieve desired rates. The dynamic inversion controller uses approximate short period and roll axis dynamics. The yaw axis controller is a sideslip rate command system. Methods are described to reduce the cross-coupling effect and maintain adequate tracking errors for control surface failures. The aerodynamic failure destabilizes the pitching moment due to angle of attack. The results show that control of the aircraft with the neural networks is easier (more damped) than without the neural networks. Simulation results show neural network augmentation of the controller improves performance with aerodynamic and control surface failures in terms of tracking error and cross-coupling reduction.

Burken, John J.↗

Combustion control using spiking neural networks

A system that controls a combustion engine stores network vectors in a memory that represent diverse and distinct spiking neural networks. The system decodes the network vectors and trains and evaluates the spiking neural networks. The system duplicates selected network vectors and crosses-over the duplicated network vectors that represent modified spiking neural networks. The system mutates the crossed-over duplicated network vectors by randomly modifying one or more portions of the crossing-over duplicated network vectors. The system meter exhaust gas into an intake manifold when an engine temperature exceeds a threshold, an engine load exceeds a threshold, an engine's rotation-per-minute rate exceeds a threshold, and a fuel flow exceeds a threshold. The system modifies fuel flow into an engine's combustion chamber on a cycle-to-cycle basis by the trained spiking neural network.

Puente, Brian P. Maldonado↗

Complex multicomponent spectrum analysis with Deep Neural Network

Here, in this paper, we present the use of deep neural networks to estimate physical parameters from complex optical emission spectra of the D β /H β transition. Specifically, we focus on estimating the radio frequency electric field vector of the lower hybrid wave and isotope ratio within the scrape-off-layer plasma of the WEST tokamak. Fitting the spectral data using a traditional non-linear least squares analysis requires many free parameters and is computationally expensive, rendering the data unusable for real-time control. By implementing relatively small neural networks, the physical parameters can be directly extracted from the spectral data with reasonable accuracy in a few milliseconds. The deep neural network prediction can serve as input for a reduced model using least-squares fitting or for real-time control. We show that deep neural networks can be an effective tool for analyzing complex multicomponent spectra, providing a speedup of more than 10 5 times compared to least residual analysis, with an accuracy of 0.5% for the isotope ratio, and 0.09 kV/cm and 0.38 kV/cm for the RF radial and poloidal electric field respectively.

47 OTHER INSTRUMENTATION↗

Improving materials property predictions for graph neural networks with minimal feature engineering *

Graph neural networks (GNNs) have been employed in materials research to predict physical and functional properties, and have achieved superior performance in several application domains over prior machine learning approaches. Recent studies incorporate features of increasing complexity such as Gaussian radial functions, plane wave functions, and angular terms to augment the neural network models, with the expectation that these features are critical for achieving a high performance. Here, we propose a GNN that adopts edge convolution where hidden edge features evolve during training and extensive attention mechanisms, and operates on simple graphs with atoms as nodes and distances between them as edges. As a result, the same model can be used for very different tasks as no other domain-specific features are used. With a model that uses no feature engineering, we achieve performance comparable with state-of-the-art models with elaborate features for formation energy and band gap prediction with standard benchmarks; we achieve even better performance when the dataset size increases. Although some domain-specific datasets still require hand-crafted features to achieve state-of-the-art results, our selected architecture choices greatly reduce the need for elaborate feature engineering and still maintain predictive power in comparison.

42 ENGINEERING↗

B-jet and c-jet identification with Neural Networks as well as combination of multivariate analyses for the search for of multivariate analyses for the search for single top-quark production

In the first part of this diploma thesis, the current version of the KIT Flavor Separator, a neural network which is able to distinguish between tagged b-quark jets and tagged c/light-quark jets, is presented. In comparison with previous versions four new input variables are utilized and new Monte Carlo samples with a larger number of simulated events are used for the training of the neural network. It is illustrated that the output of the neural network is continuously distributed between 1 and -1, whereas b-quark jets accumulate at 1, however, c-quark jets and light-quark jets have outputs next to -1. To ensure that the network output describes observed events correctly, the shapes of all input variables are compared in simulation and data. Thus the mismodelling of any input variable is excluded. Moreover, the b jet and light jet output distributions are compared with the output of samples of observed events, which are enhanced in the particular flavor. In contrast to previous versions, no b-jet output correction function has to be calculated, because the agreement between simulation and collision data is excellent for b-quark jets. For the light-jet output, correction functions are developed. Different applications of the KIT Flavor Separator are mentioned. For example it provides a precious input to all three CDF single top quark analyses. Furthermore, it is shown that the KIT Flavor Separator is a universal tool, which can be used in every high-p{sub T} analysis that requires the identification of b-quark jets with high efficiency. As it is pointed out, a further application is the estimation of the flavor composition of a given sample of observed events. In addition a neural network, which is able to separate c-quark jets from light-quark jets, is trained. It is shown, that all three flavors can be separated in the c-net-Flavor Separator plane. As a result, the uncertainties on the estimation of the flavor composition in events with one tagged jet are cut into half. In the second part of this diploma thesis, a method for the combination of three multivariate single-top analyses using an integrated luminosity of 2.2 fb{sup -1} is presented. For this purpose the discriminants of the Likelihood Function analysis, the Matrix Element method and the Neural Network analysis are used as input variables to a neural network. Overall four different networks are trained, one for events with two or three jets and one or two SecVtx tags, respectively. Using a binned likelihood function, the outputs of these networks are fitted to the output distribution of observed events. A single top-quark production cross section of {sigma}{sub single-top} = 2.2{sub -0.7}{sup +0.8} pb is measured. Ensemble tests are performed for the calculation of the sensitivity and observed significance, which are found to be 4.8{sigma} and 3.9{sigma}, respectively. Hence the improvement of this combination is roughly 8% in comparison with sensitivities found by the individual analyses. Due to the proportionality of {sigma}{sub single-top} and |V{sub tb}|{sup 2} and under the assumption V{sub tb} >> V{sub ts}, V{sub td}, a value for |V{sub tb}| is quoted: |V{sub tb}| = 0.88{sub -0.12}{sup +0.14}(exp.) {+-} 0.07(theo.). It can be seen, that the given uncertainties are too large for a verification or falsification of the unitarity assumption of the CKM-matrix. Parallel to this combination a further combination method (NEAT-combination) has been developed. This combination uses a neural network trained with a neuroevolution technique, which optimizes the neural network architecture and weights through the use of genetic algorithms. In this analysis an improvement of roughly 12% could be reached. In figure 7.1 the current situation for the measurement of the single top-quark production cross section is summarized. After collecting more data, CDF will be able to observe single top-quark production with a significance larger than 5.0{sigma}. Nevertheless, the cross section measurement will still have large uncertainties on the level of 20%. Precise measurements on the few percent level will only be possible at the LHC. Recent studies for the CMS detector showed, that the t-channel cross section can be measured with an accuracy of 7% in 1 fb{sup -1} of LHC data.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

A neural network prototyping package within IRAF

We outline our plans for incorporating a Neural Network Prototyping Package into the IRAF environment. The package we are developing will allow the user to choose between different types of networks and to specify the details of the particular architecture chosen. Neural networks consist of a highly interconnected set of simple processing units. The strengths of the connections between units are determined by weights which are adaptively set as the network 'learns'. In some cases, learning can be a separate phase of the user cycle of the network while in other cases the network learns continuously. Neural networks have been found to be very useful in pattern recognition and image processing applications. They can form very general 'decision boundaries' to differentiate between objects in pattern space and they can be used for associative recall of patterns based on partial cures and for adaptive filtering. We discuss the different architectures we plan to use and give examples of what they can do.

Bazell, D.↗

Challenges in Markov Chain Monte Carlo for Bayesian Neural Networks

Markov chain Monte Carlo (MCMC) methods have not been broadly adopted in Bayesian neural networks (BNNs). This paper initially reviews the main challenges in sampling from the parameter posterior of a neural network via MCMC. Such challenges culminate to lack of convergence to the parameter posterior. Nevertheless, this paper shows that a nonconverged Markov chain, generated via MCMC sampling from the parameter space of a neural network, can yield via Bayesian marginalization a valuable posterior predictive distribution of the output of the neural network. Further, classification examples based on multilayer perceptrons showcase highly accurate posterior predictive distributions. The postulate of limited scope for MCMC developments in BNNs is partially valid; an asymptotically exact parameter posterior seems less plausible, yet an accurate posterior predictive distribution is a tenable research avenue.

97 MATHEMATICS AND COMPUTING↗

ChemNet: A Transferable and Generalizable Deep Neural Network for Small-Molecule Property Prediction

With access to large datasets, deep neural networks through representation learning have been able to identify patterns from raw data, achieving human-level accuracy in image and speech recognition tasks. However, in chemistry, availability of large standardized and labelled datasets is scarce, and with a multitude of chemical properties of interest, chemical data is inherently small and fragmented. In this work, we explore transfer learning techniques in conjunction with the existing Chemception CNN model, to create a transferable and generalizable deep neural network for small-molecule property prediction. Our latest model, ChemNet learns in a semi-supervised manner from inexpensive labels computed from the ChEMBL database. When fine-tuned to the Tox21, HIV and FreeSolv dataset, which are 3 separate chemical tasks that ChemNet was not originally trained on, we demonstrate that ChemNet exceeds the performance of existing Chemception models, contemporary MLP models that trains on molecular fingerprints, and it matches the performance of the ConvGraph algorithm, the current state-of-the-art. Furthermore, as ChemNet has been pre-trained on a large diverse chemical database, it can be used as a universal “plug-and-play” deep neural network, which accelerates the deployment of deep neural networks for the prediction of novel small-molecule chemical properties.

Artificial Intelligence↗

PANNA: Properties from Artificial Neural Network Architectures

We report prediction of material properties from first principles is often a computationally expensive task. Recently, artificial neural networks and other machine learning approaches have been successfully employed to obtain accurate models at a low computational cost by leveraging existing example data. Here, we present a software package “Properties from Artificial Neural Network Architectures” (PANNA) that provides a comprehensive toolkit for creating neural network models for atomistic systems following the Behler–Parrinello topology. Besides the core routines for neural network training, it includes data parser, descriptor builder for Behler–Parrinello class of symmetry functions and force-field generator suitable for integration within molecular dynamics packages. PANNA offers a variety of activation and cost functions, regularization methods, as well as the possibility of using fully-connected networks with custom size for each atomic species. PANNA benefits from the optimization and hardware-flexibility of the underlying TensorFlow engine which allows it to be used on multiple CPU/GPU/TPU systems, making it possible to develop and optimize neural network models based on large datasets.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Antenna analysis using neural networks

Conventional computing schemes have long been used to analyze problems in electromagnetics (EM). The vast majority of EM applications require computationally intensive algorithms involving numerical integration and solutions to large systems of equations. The feasibility of using neural network computing algorithms for antenna analysis is investigated. The ultimate goal is to use a trained neural network algorithm to reduce the computational demands of existing reflector surface error compensation techniques. Neural networks are computational algorithms based on neurobiological systems. Neural nets consist of massively parallel interconnected nonlinear computational elements. They are often employed in pattern recognition and image processing problems. Recently, neural network analysis has been applied in the electromagnetics area for the design of frequency selective surfaces and beam forming networks. The backpropagation training algorithm was employed to simulate classical antenna array synthesis techniques. The Woodward-Lawson (W-L) and Dolph-Chebyshev (D-C) array pattern synthesis techniques were used to train the neural network. The inputs to the network were samples of the desired synthesis pattern. The outputs are the array element excitations required to synthesize the desired pattern. Once trained, the network is used to simulate the W-L or D-C techniques. Various sector patterns and cosecant-type patterns (27 total) generated using W-L synthesis were used to train the network. Desired pattern samples were then fed to the neural network. The outputs of the network were the simulated W-L excitations. A 20 element linear array was used. There were 41 input pattern samples with 40 output excitations (20 real parts, 20 imaginary). A comparison between the simulated and actual W-L techniques is shown for a triangular-shaped pattern. Dolph-Chebyshev is a different class of synthesis technique in that D-C is used for side lobe control as opposed to pattern shaping. The interesting thing about D-C synthesis is that the side lobes have the same amplitude. Five-element arrays were used. Again, 41 pattern samples were used for the input. Nine actual D-C patterns ranging from -10 dB to -30 dB side lobe levels were used to train the network. A comparison between simulated and actual D-C techniques for a pattern with -22 dB side lobe level is shown. The goal for this research was to evaluate the performance of neural network computing with antennas. Future applications will employ the backpropagation training algorithm to drastically reduce the computational complexity involved in performing EM compensation for surface errors in large space reflector antennas.

Smith, William T.↗

Using an Extended Kalman Filter Learning Algorithm for Feed-Forward Neural Networks to Describe Tracer Correlations

In this study a new extended Kalman filter (EKF) learning algorithm for feed-forward neural networks (FFN) is used. With the EKF approach, the training of the FFN can be seen as state estimation for a non-linear stationary process. The EKF method gives excellent convergence performances provided that there is enough computer core memory and that the machine precision is high. Neural networks are ideally suited to describe the spatial and temporal dependence of tracer-tracer correlations. The neural network performs well even in regions where the correlations are less compact and normally a family of correlation curves would be required. For example, the CH4-N2O correlation can be well described using a neural network trained with the latitude, pressure, time of year, and CH4 volume mixing ratio (v.m.r.). The neural network was able to reproduce the CH4-N2O correlation with a correlation coefficient between simulated and training values of 0.9997. The neural network Fortran code used is available for download.

Lary, David J.↗

A neural network approach to cloud classification

It is shown that, using high-spatial-resolution data, very high cloud classification accuracies can be obtained with a neural network approach. A texture-based neural network classifier using only single-channel visible Landsat MSS imagery achieves an overall cloud identification accuracy of 93 percent. Cirrus can be distinguished from boundary layer cloudiness with an accuracy of 96 percent, without the use of an infrared channel. Stratocumulus is retrieved with an accuracy of 92 percent, cumulus at 90 percent. The use of the neural network does not improve cirrus classification accuracy. Rather, its main effect is in the improved separation between stratocumulus and cumulus cloudiness. While most cloud classification algorithms rely on linear parametric schemes, the present study is based on a nonlinear, nonparametric four-layer neural network approach. A three-layer neural network architecture, the nonparametric K-nearest neighbor approach, and the linear stepwise discriminant analysis procedure are compared. A significant finding is that significantly higher accuracies are attained with the nonparametric approaches using only 20 percent of the database as training data, compared to 67 percent of the database in the linear approach.

Lee, Jonathan↗

Experiments in Neural-Network Control of a Free-Flying Space Robot

Four important generic issues are identified and addressed in some depth in this thesis as part of the development of an adaptive neural network based control system for an experimental free flying space robot prototype. The first issue concerns the importance of true system level design of the control system. A new hybrid strategy is developed here, in depth, for the beneficial integration of neural networks into the total control system. A second important issue in neural network control concerns incorporating a priori knowledge into the neural network. In many applications, it is possible to get a reasonably accurate controller using conventional means. If this prior information is used purposefully to provide a starting point for the optimizing capabilities of the neural network, it can provide much faster initial learning. In a step towards addressing this issue, a new generic Fully Connected Architecture (FCA) is developed for use with backpropagation. A third issue is that neural networks are commonly trained using a gradient based optimization method such as backpropagation; but many real world systems have Discrete Valued Functions (DVFs) that do not permit gradient based optimization. One example is the on-off thrusters that are common on spacecraft. A new technique is developed here that now extends backpropagation learning for use with DVFs. The fourth issue is that the speed of adaptation is often a limiting factor in the implementation of a neural network control system. This issue has been strongly resolved in the research by drawing on the above new contributions.

NEURAL NETS↗

Making Phase-Picking Neural Networks More Consistent and Interpretable

Improving the interpretability of phase-picking neural networks remains an important task to facilitate their deployment to routine, real-time seismic monitoring. The popular phase-picking neural networks published in the literature lack interpretability because their output prediction scores do not necessarily correspond with the reliability of phase picks and can even be highly inconsistent depending on how we window the waveform data. Here, we show that systematically shifting the waveforms during training and using an antialiasing filter within the neural network architecture can substantially improve the consistency of the output prediction scores and can even make them scale with the signal-to-noise ratios of the waveforms. We demonstrate the improvements by applying these approaches to a commonly used phase-picking neural network architecture and using waveform data from the 2019 Ridgecrest earthquake sequence.

58 GEOSCIENCES↗

Towards a robust out-of-the-box neural network model for genomic data

The accurate prediction of biological features from genomic data is paramount for precision medicine and sustainable agriculture. For decades, neural network models have been widely popular in fields like computer vision, astrophysics and targeted marketing given their prediction accuracy and their robust performance under big data settings. Yet neural network models have not made a successful transition into the medical and biological world due to the ubiquitous characteristics of biological data such as modest sample sizes, sparsity, and extreme heterogeneity. Here, we investigate the robustness, generalization potential and prediction accuracy of widely used convolutional neural network and natural language processing models with a variety of heterogeneous genomic datasets. Mainly, recurrent neural network models outperform convolutional neural network models in terms of prediction accuracy, overfitting and transferability across the datasets under study. While the perspective of a robust out-of-the-box neural network model is out of reach, we identify certain model characteristics that translate well across datasets and could serve as a baseline model for translational researchers.

59 BASIC BIOLOGICAL SCIENCES↗

Hierarchical neural networks for autonomous data analysis and decision making

A neural network based data analysis and decision making system to increase the autonomy of a planetary rover or similar exploratory vehicle is presented. A hierarchical series of neural networks for real time analysis of scientific images is used. The system under development emphasizes analysis of multispectral images by classifier and feature detector neural networks, to provide information on the mineral composition of a scene. A hierarchy of alternating analysis and decision making networks is being developed to allow increasingly fine scale analysis in regions of the image that are potentially important. It is noted that this system will facilitate both the selection of high priorty scientific information for transmission to earth, and the autonomous collection of rocks and soil for sample return.

Eberlein, Susan↗

Analytic Theory for the Dynamics of Wide Quantum Neural Networks

Here, parametrized quantum circuits can be used as quantum neural networks and have the potential to outperform their classical counterparts when trained for addressing learning problems. To date, much of the results on their performance on practical problems are heuristic in nature. In particular, the convergence rate for the training of quantum neural networks is not fully understood. Here, we analyze the dynamics of gradient descent for the training error of a class of variational quantum machine learning models. We define wide quantum neural networks as parametrized quantum circuits in the limit of a large number of qubits and variational parameters. Then, we find a simple analytic formula that captures the average behavior of their loss function and discuss the consequences of our findings. For example, for random quantum circuits, we predict and characterize an exponential decay of the residual training error as a function of the parameters of the system. Finally, we validate our analytic results with numerical experiments.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗