Search NASA⌕ Search

SEARCH · Search NASA

Results for “gradient methods”

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 937 records · Page 52

Multirobot Collaborative Pursuit Target Robot by Improved MADDPG

Policy formulation is one of the main problems in multirobot systems, especially in multirobot pursuit-evasion scenarios, where both sparse rewards and random environment changes bring great difficulties to find better strategy. Existing multirobot decision-making methods mostly use environmental rewards to promote robots to complete the target task that cannot achieve good results. This paper proposes a multirobot pursuit method based on improved multiagent deep deterministic policy gradient (MADDPG), which solves the problem of sparse rewards in multirobot pursuit-evasion scenarios by combining the intrinsic reward and the external environment. The state similarity module based on the threshold constraint is as a part of the intrinsic reward signal output by the intrinsic curiosity module, which is used to balance overexploration and insufficient exploration, so that the agent can use the intrinsic reward more effectively to learn better strategies. The simulation experiment results show that the proposed method can improve the reward value of robots and the success rate of the pursuit task significantly. The intuitive change is obviously reflected in the real-time distance between the pursuer and the escapee, the pursuer using the improved algorithm for training can get closer to the escapee more quickly, and the average following distance also decreases.

Zhou, Xiao↗

Flexible Thermoelectric Devices of Ultrahigh Power Factor by Scalable Printing and Interface Engineering

Printing is a versatile method to transform semiconducting nanoparticle inks into functional and flexible devices. In particular, thermoelectric nanoparticles are attractive building blocks to fabricate flexible devices for energy harvesting and cooling applications. However, the performance of printed devices are plagued by poor interfacial connections between nanoparticles and resulting low carrier mobility. While many rigid bulk materials have shown a thermoelectric figure of merit ZT greater than unity, it is an exacting challenge to develop flexible materials with ZT near unity. Here, a scalable screen-printing method to fabricate high-performance and flexible thermoelectric devices is reported. A tellurium-based nanosolder approach is employed to bridge the interfaces between the BiSbTe particles during the postprinting sintering process. The printed BiSbTe flexible films demonstrate an ultrahigh room-temperature power factor of 3 mW m -1 K -2 and ZT about 1, significantly higher than the best reported values for flexible films. A fully printed thermoelectric generator produces a high power density of 18.8 mW cm -2 achievable with a small temperature gradient of 80 °C. This screen-printing method, which directly transforms thermoelectric nanoparticles into high-performance and flexible devices, presents a significant leap to make thermoelectrics a commercially viable technology for a broad range of energy harvesting and cooling applications.

36 MATERIALS SCIENCE↗

Uncertainty quantification of graph convolution neural network models of evolving processes

The application of neural network models to scientific machine learning tasks has proliferated in recent years. In particular, neural networks have proved to be adept at modeling processes with spatial–temporal complexity. Nevertheless, these highly parameterized models have garnered skepticism in their ability to produce outputs with quantified error bounds over the regimes of interest. Hence there is a need to find uncertainty quantification methods that are suitable for neural networks. In this work we present comparisons of the parametric uncertainty quantification of neural networks modeling complex spatial–temporal processes with Hamiltonian Monte Carlo and Stein variational gradient descent and its projected variant. Specifically we apply these methods to graph convolutional neural network models of evolving systems modeled with recurrent neural network and neural ordinary differential equations architectures. We show that Stein variational inference is a viable alternative to Monte Carlo methods with some clear advantages for complex neural network models. For our exemplars, Stein variational interference gave similar pushed forward uncertainty profiles through time compared to Hamiltonian Monte Carlo, albeit with generally more generous variance. As a result, projected Stein variational gradient descent also produced similar uncertainty profiles to the non-projected counterpart, but large reductions in the active weight space were confounded by the stability of the neural network predictions and the convoluted likelihood landscape.

36 MATERIALS SCIENCE↗

An inexact semismooth Newton method with application to adaptive randomized sketching for dynamic optimization

In many applications, one can only access the inexact gradients and inexact hessian times vector products. Thus it is essential to consider algorithms that can handle such inexact quantities with a guaranteed convergence to solution. An inexact adaptive and provably convergent semismooth Newton method is considered to solve constrained optimization problems. In particular, dynamic optimization problems, which are known to be highly expensive, are the focus. A memory efficient semismooth Newton algorithm is introduced for these problems. The source of efficiency and inexactness is the randomized matrix sketching. Further, applications to optimization problems constrained by partial differential equations are also considered.

97 MATHEMATICS AND COMPUTING↗

Gaussian processes meet NeuralODEs: a Bayesian framework for learning the dynamics of partially observed systems from scarce and noisy data

We present a machine learning framework (GP-NODE) for Bayesian model discovery from partial, noisy and irregular observations of nonlinear dynamical systems. The proposed method takes advantage of differentiable programming to propagate gradient information through ordinary differential equation solvers and perform Bayesian inference with respect to unknown model parameters using Hamiltonian Monte Carlo sampling and Gaussian Process priors over the observed system states. This allows us to exploit temporal correlations in the observed data, and efficiently infer posterior distributions over plausible models with quantified uncertainty. The use of the Finnish Horseshoe as a sparsity-promoting prior for free model parameters also enables the discovery of parsimonious representations for the latent dynamics. A series of numerical studies is presented to demonstrate the effectiveness of the proposed GP-NODE method including predator–prey systems, systems biology and a 50-dimensional human motion dynamical system. This article is part of the theme issue ‘Data-driven prediction in dynamical systems’.

Science & Technology - Other Topics↗

Study of structural and electronic properties of ScN and ScAs in rocksalt and zincblende structure: A DFT approach

In this paper, we have studied the structural and electronic properties of ScN and ScAs in zincblende (ZB) and rocksalt (RS) phases. We have employed the full potential linearized augmented plane wave (FP-APW) method within the density functional theory (DFT). Generalized gradient approximation (GGA), due to Perdew-Burke-Ernzerhof (PBE) has been used to estimate the exchange-correlation functional. Our band structure results for ScN shows the metallic nature, whereas ScAs shows the semiconducting behavior. The obtained results are in excellent agreement with earlier reported data.

36 MATERIALS SCIENCE↗

Nonlinear electronic excitations in crystalline solids using meta-generalized gradient approximation and hybrid functional in time-dependent density functional theory

We develop methods to calculate electron dynamics in crystalline solids in real-time time-dependent density functional theory employing exchange-correlation potentials which reproduce band gap energies of dielectrics; a meta-generalized gradient approximation was proposed by Tran and Blaha [Phys. Rev. Lett. 102, 226401 (2009)] (TBm-BJ) and a hybrid functional was proposed by Heyd, Scuseria, and Ernzerhof [J. Chem. Phys. 118, 8207 (2003)] (HSE). In time evolution calculations employing the TB-mBJ potential, we have found it necessary to adopt the predictor-corrector step for a stable time evolution. We have developed a method to evaluate electronic excitation energy without referring to the energy functional which is unknown for the TB-mBJ potential. For the HSE functional, we have developed a method for the operation of the Fock-like term in Fourier space to facilitate efficient use of massive parallel computers equipped with graphic processing units. We compare electronic excitations in silicon and germanium induced by femtosecond laser pulses using the TB-mBJ, HSE, and a simple local density approximation (LDA). At low laser intensities, electronic excitations are found to be sensitive to the band gap energy: they are close to each other using TB-mBJ and HSE and are much smaller in LDA. At high laser intensities close to the damage threshold, electronic excitation energies do not differ much among the three cases.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Approximating Nash Equilibrium in Day-ahead Electricity Market Bidding with Multi-agent Deep Reinforcement Learning

In this paper, a day-ahead electricity market bidding problem with multiple strategic generation company (GEN-CO) bidders is studied. The problem is formulated as a Markov game model, where GENCO bidders interact with each other todevelop their optimal day-ahead bidding strategies. Considering unobservable information in the problem, a model-free and data-driven approach, known as multi-agent deep deterministic policy gradient (MADDPG), is applied for approximating the Nash equilibrium (NE) in the above Markov game. The MADDPG algorithm has the advantage of generalization due to the automatic feature extraction ability of the deep neural networks. The algorithm is tested on an IEEE 30-bus system with three competitive GENCO bidders in both an uncongested caseand a congested case. Comparisons with a truthful bidding strategy and state-of-the-art deep reinforcement learning methods including deep Q network and deep deterministic policy gradient (DDPG) demonstrate that the applied MADDPG algorithm can find a superior bidding strategy for all the market participants with increased profit gains. In addition, the comparison with a conventional model-based method shows that the MADDPG algorithm has higher computational efficiency, which is feasible for real-world applications.

24 POWER TRANSMISSION AND DISTRIBUTION↗

A Gradient Based Subgrid-Scale Parameterization for Ocean Mesoscale Eddies

Mesoscale eddies play an important role in transport of heat and biogeochemical tracers in the global ocean circulation. Resolving these energetic eddies, however, is challenging in ocean general circulation models (OGCM) because it requires a horizontal grid spacing of ≲1/8° that is computationally expensive. As a result, we are required to parameterize mesoscale eddy effects on large-scale ocean flows. In this work, we introduce a new subgrid-scale (SGS) model that is developed based on a Taylor series expansion of resolved variables to parameterize subgrid mesoscale eddy transports and momentum fluxes in OGCM. We have performed an a priori study to evaluate the performance of our new gradient model using high-resolution ocean simulations. Our results show that the gradient model well predicts the actual SGS thickness fluxes in the zonal and meridional directions in coarse-resolution simulations with the grid spacing ≳1/4°. The unresolved kinetic energy at the ocean surface is also skillfully estimated. More importantly, unlike current mesoscale eddy parameterizations, which are mainly developed based on an assumption of flat bottom ocean, our new SGS model can capture the structure of unresolved standing meanders at the ocean surface. We have also developed a dynamic procedure for setting in non-dimensional parameters in our new parameterization through a non-ad hoc and tuning-free method. Overall, this work suggests that implementing the gradient model in OGCM can improve the model accuracy with an affordable computational cost in eddy-permitting and non-eddying simulations.

54 ENVIRONMENTAL SCIENCES↗

{sup 1}D states of the beryllium atom: Quantum mechanical nonrelativistic calculations employing explicitly correlated Gaussian functions

Very accurate finite-nuclear-mass variational nonrelativistic calculations are performed for the lowest five {sup 1}D states (1s{sup 2} 2p{sup 2}, 1s{sup 2} 2s{sup 1} 3d{sup 1}, 1s{sup 2} 2s{sup 1} 4d{sup 1}, 1s{sup 2} 2s{sup 1} 5d{sup 1}, and 1s{sup 2} 2s{sup 1} 6d{sup 1}) of the beryllium atom ({sup 9}Be). The wave functions of the states are expanded in terms of all-electron explicitly correlated Gaussian functions. The exponential parameters of the Gaussians are optimized using the variational method with the aid of the analytical energy gradient determined with respect to those parameters. The calculations exemplify the level of accuracy that is now possible with Gaussians in describing bound states of a four-electron system where some of the electrons are excited into higher angular states.

74 ATOMIC AND MOLECULAR PHYSICS↗

Turbulence model reduction by deep learning

A defining problem of turbulence theory is to produce a predictive model for turbulent fluxes. These have profound implications for virtually all aspects of the turbulence dynamics. In magnetic confinement devices, drift-wave turbulence produces anomalous fluxes via cross-correlations between fluctuations. In this work, we introduce an alternative, data-driven method for parametrizing these fluxes. The method uses deep supervised learning to infer a reduced mean-field model from a set of numerical simulations. We apply the method to a simple drift-wave turbulence system and find a significant new effect which couples the particle flux to the local gradient of vorticity. Notably, here, this effect is much stronger than the oft-invoked shear suppression effect. We also recover the result via a simple calculation. The vorticity gradient effect tends to modulate the density profile. In addition, our method recovers a model for spontaneous zonal flow generation by negative viscosity, stabilized by nonlinear and hyperviscous terms. We highlight the important role of symmetry to implementation of the new method.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

New Method for Measuring the Anchoring Energy of Strongly-Bound Membrane-Associated Proteins [Method for measuring the anchoring energy of strongly-bound membrane-associated proteins].

Here, we describe a new method to measure the activation energy required to remove a strongly-bound membrane-associated protein from a lipid membrane (anchoring energy). It is based on measuring the rate of release of a liposome-bound protein during centrifugation on a sucrose gradient as a function of time and temperature. The method was used to determine anchoring energy for the soluble dengue virus envelope protein (sE) strongly bound to 80:20 POPC:POPG liposomes at pH 5.5. We also measured the binding energy of sE at the same pH for the initial, predominantly reversible, phase of binding to a 70:30 PC:PG lipid bilayer. The anchoring energy (37 +/- 1.7 kcal/mol, 20% PG) was found to be much larger than the binding energy (7.8 +/- 0.3 kcal/mol for 30% PG, or est. 7.0 kcal/mol for 20% PG). This is consistent with data showing that free sE is a monomer at pH 5.5, but assembles into trimers after associating with membranes. But, trimerization alone is insufficient to account for the observed difference in energies, and we conclude that some energy dissipation occurs during the release process. This new method to determine anchoring energy should be useful to understand the complex interactions of integral monotopic proteins and strongly-bound peripheral membrane proteins with lipid membranes.

59 BASIC BIOLOGICAL SCIENCES↗

Experimentally studied dynamic dose interplay does not meaningfully affect target dose in VMAT SBRT lung treatments

Purpose: The effects of respiratory motion on the tumor dose can be divided into the gradient and interplay effects. While the interplay effect is likely to average out over a large number of fractions, it may play a role in hypofractionated [stereotactic body radiation therapy (SBRT)] treatments. This subject has been extensively studied for intensity modulated radiation therapy but less so for volumetric modulated arc therapy (VMAT), particularly in application to hypofractionated regimens. Also, no experimental study has provided full four-dimensional (4D) dose reconstruction in this scenario. The authors demonstrate how a recently described motion perturbation method, with full 4D dose reconstruction, is applied to describe the gradient and interplay effects during VMAT lung SBRT treatments.Methods: VMAT dose delivered to a moving target in a patient can be reconstructed by applying perturbations to the treatment planning system-calculated static 3D dose. Ten SBRT patients treated with 6 MV VMAT beams in five fractions were selected. The target motion (motion kernel) was approximated by 3D rigid body translation, with the tumor centroids defined on the ten phases of the 4DCT. The motion was assumed to be periodic, with the period T being an average from the empirical 4DCT respiratory trace. The real observed tumor motion (total displacement ≤8 mm) was evaluated first. Then, the motion range was artificially increased to 2 or 3 cm. Finally, T was increased to 60 s. While not realistic, making T comparable to the delivery time elucidates if the interplay effect can be observed. For a single fraction, the authors quantified the interplay effect as the maximum difference in the target dosimetric indices, most importantly the near-minimum dose (D{sub 99%}), between all possible starting phases. For the three- and five-fractions, statistical simulations were performed when substantial interplay was found.Results: For the motion amplitudes and periods obtained from the 4DCT, the interplay effect is negligible (<0.2%). It is also small (0.9% average, 2.2% maximum) when the target excursion increased to 2–3 cm. Only with large motion and increased period (60 s) was a significant interplay effect observed, with D{sub 99%} ranging from 16% low to 17% high. The interplay effect was statistically significantly lower for the three- and five-fraction statistical simulations. Overall, the gradient effect dominates the clinical situation.Conclusions: A novel method was used to reconstruct the volumetric dose to a moving tumor during lung SBRT VMAT deliveries. With the studied planning and treatment technique for realistic motion periods, regardless of the amplitude, the interplay has nearly no impact on the near-minimum dose. The interplay effect was observed, for study purposes only, with the period comparable to the VMAT delivery time.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

The first principle investigations of structural and electronic properties of ZnTe

Structural and electronic properties of ZnTe are investigated within the framework of density functional theory. The pseudopotential method along with local density approximation and generalized gradient approximation for the exchange-correlation potential is used within the quantum espresso package. The optimized equilibrium lattice parameter, bulk modulus and its pressure derivative are determined. The electronic band structure, density of states and partial density of states are calculated. The results are compared with available theoretical calculations and experimental results.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

The Thick Level-Set model for dynamic fragmentation

The Thick Level-Set (TLS) model is implemented to simulate brittle media undergoing dynamic fragmentation. This non-local model is discretized by the finite element method with damage represented as a continuous field over the domain. A level-set function defines the extent and severity of damage, and a length scale is introduced to limit the damage gradient. Numerical studies in one dimension demonstrate that the proposed method reproduces the rate-dependent energy dissipation and fragment length observations from analytical, numerical, and experimental approaches. In conclusion, additional studies emphasize the importance of appropriate bulk constitutive models and sufficient spatial resolution of the length scale.

97 MATHEMATICS AND COMPUTING↗

Pressure distribution along the AGS vacuum chambers with new types of pump out conduits

The AGS HEBT and ring vacuum system is monitored by the discharge current of the magnet ion pumps, which is proportional to the pressure at the inlet port of these ion pumps. The discharge current is measured and suitably calibrated to indicate the ion pump pressure. In order to calculate the vacuum chamber pressure from the ion pump pressure, a detailed analysis is essential to compute their difference in different scenarios. Such analysis has been carried out numerically in the past for the system with the older type of pump out conduits, and similar analysis using FEM in ANSYS is presented in this paper with the newer type of pump out conduit.

43 PARTICLE ACCELERATORS↗

NMR, MRI, and spectroscopic MRI in inhomogeneous fields

A method for locally creating effectively homogeneous or "clean" magnetic field gradients (of high uniformity) for imaging (with NMR, MRI, or spectroscopic MRI) both in in-situ and ex-situ systems with high degrees of inhomogeneous field strength. THe method of imaging comprises: a) providing a functional approximation of an inhomogeneous static magnetic field strength B.sub.0({right arrow over (r)}) at a spatial position {right arrow over (r)}; b) providing a temporal functional approximation of {right arrow over (G)}.sub.shim(t) with i basis functions and j variables for each basis function, resulting in v.sub.ij variables; c) providing a measured value .OMEGA., which is an temporally accumulated dephasing due to the inhomogeneities of B.sub.0({right arrow over(r)}); and d) minimizing a difference in the local dephasing angle .phi.({right arrow over (r)},t)=.gamma..intg..sub.0.sup.t{square root over (|{right arrow over (B)}.sub.1({right arrow over (r)},t')|.sup.2+({right arrow over (r)}{right arrow over (G)}.sub.shimG.sub.shim(t')+.parallel.{right arrow over (B)}.sub.0({right arrow over (r)}).parallel..DELTA..omega.({right arrow over (r)},t'/.gamma/).sup.2)}dt'-.OMEGA. by varying the v.sub.ij variables to form a set of minimized v.sub.ij variables. The method requires calibration of the static fields prior to minimization, but may thereafter be implemented without such calibration, may be used in open or closed systems, and potentially portable systems.

47 OTHER INSTRUMENTATION↗

Multi-scale fusion algorithm of intensity and polarization-difference images based on edge information enhancement

To better integrate complementary and redundant information from different source images, improve the edge information, and facilitate the target detection. A multi-scale fusion algorithm of intensity and polarization-difference (PD) images based on edge information enhancement is proposed. Firstly, intensity images are obtained by the polarization information analysis method. PD images are obtained by the adaptive polarization-difference imaging approach based on the principle of minimum mutual information. Secondly, guided filter, affine transformations and Block-Matching and 3D filtering are embedded in visibility enhancement to improve the intensity and PD images. Thirdly, the two images are decomposed into high-frequency and low-frequency images by the dual-tree complex wavelet transform (DT-CWT). The high-frequency and low-frequency images are fused by the fusion rules based on edge detection and the regional variance matching degree respectively. Finally, the fusion image is obtained by the inverse DT-CWT. Experimental results demonstrate that fusion images of the proposed algorithm are significantly improved in information entropy, average gradient, and spatial frequency. Compared with the existing methods, it can achieve a better edge enhancement for images in a turbid medium.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗