Search NASA⌕ Search

SEARCH · Search NASA

Results for “Directional Gaussian smoothing”

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 19 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

Convergence analysis for a nonlocal gradient descent method via directional Gaussian smoothing

We analyze the convergence of a nonlocal gradient descent method for minimizing a class of high-dimensional non-convex functions, where a directional Gaussian smoothing (DGS) is proposed to define the nonlocal gradient (also referred to as the DGS gradient). The method was first proposed in [Zhang et al., Enabling long-range exploration in minimization of multimodal functions, UAI 2021], in which multiple numerical experiments showed that replacing the traditional local gradient with the DGS gradient can help the optimizers escape local minima more easily and significantly improve their performance. However, a rigorous theory for the efficiency of the method on nonconvex landscape is lacking. In this work, we investigate the scenario where the objective function is composed of a convex function, perturbed by deterministic oscillating noise. We provide a convergence theory under which the iterates exponentially converge to a tightened neighborhood of the solution, whose size is characterized by the noise wavelength. Here, we also establish a correlation between the optimal values of the Gaussian smoothing radius and the noise wavelength, thus justifying the advantage of using moderate or large smoothing radii with the method. Furthermore, if the noise level decays to zero when approaching the global minimum, we prove that DGS-based optimization converges to the exact global minimum with linear rates, similarly to standard gradient-based methods in optimizing convex functions. Several numerical experiments are provided to confirm our theory and illustrate the superiority of the approach over those based on the local gradient.

Tran, Hoang [Oak Ridge National Laboratory (ORNL),↗
DOE OSTI2021

Enabling Long-range Exploration in Minimization of Multimodal Functions

We consider the problem of minimizing multi-modal loss functions with a large number of local optima. Since the local gradient points to the direction of the steepest slope in an infinitesimal neighborhood, an optimizer guided by the local gradient is often trapped in a local minimum. To address this issue, we develop a novel nonlocal gradient to skip small local minima by capturing major structures of the loss’s landscape in black-box optimization. The nonlocal gradient is defined by a directional Gaussian smoothing (DGS) approach. The key idea of DGS is to conducts 1D long-range exploration with a large smoothing radius along d orthogonal directions in Rd, each of which defines a nonlocal directional derivative as a 1D integral. Such long-range exploration enables the nonlocal gradient to skip small local minima. The d directional derivatives are then assembled to form the nonlocal gradient. We use the Gauss-Hermite quadrature rule to approximate the d 1D integrals to obtain an accurate estimator. The superior performance of our method is demonstrated in three sets of examples, including benchmark functions for global optimization, and two real-world scientific problems.

Zhang, Jiaxin↗
DOE OSTI2024

Material Model Parameters Optimization in Liquid Mercury Target Dynamics Simulation With Machine Learning Surrogates

A pulsed spallation target is subjected to very short (∼0.7μs) but intense loads (23.3 kJ) from repeated proton pulses, which knock away neutrons from the mercury atoms’ nuclei for a wide range application in physics, engineering, medicine, petroleum exploration, biology, chemistry, etc. The effect of this pulsed loading on the stainless-steel target module which contains the flowing mercury target material is difficult to predict not only due to its short but intense explosive-like physical reaction, but also the nonlinear material behavior of the liquid mercury in the structure. Injecting small helium bubbles in the mercury has been an efficient method of mitigating the pressure wave at high power level stage. However, prediction of the resultant loading on the target is more difficult when helium gas is intentionally injected into the mercury. A 2-phase material model that incorporates the Rayleigh-Plesset (R-P) model is expected to address this complex multi-physics dynamics problem by including the bubble dynamics in the liquid mercury. A parameter sensitivity study was firstly employed to understand their impact on the simulation strains. The investigated parameters included E, μ, γ, σ, n, VFgas, and gas cumulative volume curve control parameters a and b. Verification and validation results from sparse polynomial expansions (SPE) method and directional Gaussian smoothing (DGS) optimization show that the surrogate model had training error of ∼7% and validation error of ∼15%, indicating that machine learning methods and surrogate models can help optimize the uncertain parameters in the complex 2-phase material model. This approach is expected to fill the knowledge gap between unknown liquid-gas mixture material model and measured vessel strain responses.

Lin, Lianshan↗
DOE OSTI2020

A Nonlocal-Gradient Descent Method for Inverse Design in Nanophotonics

Local-gradient-based optimization approaches lack nonlocal exploration abilityrequired for escaping from local minima when searching non-convex landscapes.A directional Gaussian smoothing (DGS) approach was recently proposed in [29]and used to define a truly nonlocal gradient, referred to as the DGS gradient, inorder to enable nonlocal exploration in high-dimensional black-box optimization.Promising results show that replacing the traditional local gradient with the nonlocalDGS gradient can significantly improve the performance of gradient-based methodsin optimizing highly multi-modal loss functions. However, the current DGS methodis designed for unbounded and uncontrained optimization problems, making itinapplicable to real-world engineering optimization problems where the tuningparameters are often bounded and the loss function is usually constrained byphysical processes. In this work, we propose to extend to the DGS approachto the constrained inverse design framework in order to find better optima ofmulti-modal loss functions. A series of adaptive strategies for smoothing radiusand learning rate updating are developed to improve the computational efficiencyand robustness. Our methodology is demonstrated by an example of designing ananoscale wavelength demultiplexer, and shows superior performance compared tothe state-of-the-art approaches. By incorporating volume constraints, the optimizeddesign achieves an equivalently high performance but significantly reduces theamount of material usage.

Bi, Sirui↗
NASA NTRS1993

Improving the reconstruction of the velocity potential and primordial density fluctuations by choice of smoothing windows

Recently considerable attention has been focused on improving algorithms for restoring primordial density fluctuations in the universe or investigating large-scale velocity fields by going to higher order approximations relating nonlinear states to their initial conditions. An alternative approach is to investigate the use of alternative smoothing windows within the context of first-order approximations. I present evidence that the universally used Gaussian smoothing window is far from optimal. A sharp truncation P(k) = 0 for k greater than k(c) leads to a much more direct connection to initial conditions for either 'Gaussianization' methods or use of the Zel'dovich approximation, as discussed in the text.

Melott, Adrian L.↗
NASA NTRS1993

Controlled experiments in cosmological gravitational clustering

A systematic study is conducted of gravitational instability in 3D on the basis of power-law initial spectra with and without spectral cutoff, emphasizing nonlinear effects and measures of nonlinearity; effects due to short and long waves in the initial conditions are separated. The existence of second-general pancakes is confirmed, and it is noted that while these are inhomogeneous, they generate a visually strong signal of filamentarity. An explicit comparison of smoothed initial conditions with smoothed envelope models also reconfirms the need to smooth over a scale larger than any nonlinearity, in order to extrapolate directly by linear theory from Gaussian initial conditions.

Melott, Adrian L.↗
NASA NTRS1986

Theory for computing the field scattered from a smooth inflected surface

A theory is described for computing the reflected or scattered field from a smooth body with inflection points. These inflections occur in certain directions at each surface point for which the total (Gaussian) curvature is zero or negative. For surface illumination in one of these critical directions, the usual reflection formulas obtained by the high-frequency approximation are inapplicable, and a shadow zone exists in the reflected field. Scattering into the shadow zone is treated, as well as specular reflection. This theory should have a variety of applications such as for certain optics problems, computer graphics modeling of three-dimensional shapes, and the design and analysis of specialized microwave reflector antennas.

Barger, R. L.↗
NASA NTRS1989

Analysis of high-incidence separated flow past airfoils

An unsteady Navier-Stokes (NS) analysis is developed and used to carefully examine high-incidence aerodynamic separated flows past airfoils. Clustered conformal C-grids are employed for the 12 percent thick symmetric Joukowski airfoil as well as for the NACA 0012 airfoil with a sharp trailing edge. The clustering is controlled by appropriate one-dimensional stretching transformations. An attempt is made to resolve many of the dominant scales of an unsteady flow with massive separation, while maintaining the transformation metrics to be smooth and continuous in the entire flow field. A fully implicit time-marching alternating-direction implicit-block Gaussian elimination (ADI-BGE) method is employed, in which no use is made of any explicit artificial dissipation. Detailed results are obtained for massively separated, unsteady flow past symmetric Joukowski and NACA 0012 airfoils.

Chia, K. N.↗
DOE OSTI2023

BeyondPlanck: V. Minimal ADC Corrections for Planck LFI

We describe the correction procedure for Analog-to-Digital Converter (ADC) differential non-linearities (DNL) adopted in the Bayesian end-to-end BEYONDPLANCK analysis framework. This method is nearly identical to that developed for the official Planck Low Frequency Instrument (LFI) Data Processing Center (DPC) analysis, and relies on the binned rms noise profile of each detector data stream. However, rather than building the correction profile directly from the raw rms profile, we first fit a Gaussian to each significant ADC-induced rms decrement, and then derive the corresponding correction model from this smooth model. The main advantage of this approach is that only samples which are significantly affected by ADC DNLs are corrected, as opposed to the DPC approach in which the correction is applied to all samples, filtering out signals not associated with ADC DNLs. The new corrections are only applied to data for which there is a clear detection of the non-linearities, and for which they perform at least comparably with the DPC corrections. Out of a total of 88 LFI data streams (sky and reference load for each of the 44 detectors) we apply the new minimal ADC corrections in 25 cases, and maintain the DPC corrections in 8 cases. All these corrections are applied to 44 or 70 GHz channels, while, as in previous analyses, none of the 30 GHz ADCs show significant evidence of non-linearity. By comparing the BEYONDPLANCK and DPC ADC correction methods, we estimate that the residual ADC uncertainty is about two orders of magnitude below the total noise of both the 44 and 70 GHz channels, and their impact on current cosmological parameter estimation is small. However, we also show that non-idealities in the ADC corrections can generate sharp stripes in the final frequency maps, and these could be important for future joint analyses with the Planck High Frequency Instrument (HFI), Wilkinson Microwave Anisotropy Probe (WMAP), or other datasets. We therefore conclude that, although the existing corrections are adequate for LFI-based cosmological parameter analysis, further work on LFI ADC corrections is still warranted.

79 ASTRONOMY AND ASTROPHYSICS↗
NASA NTRS1986

Partial and Interaction Spline Models for the Semiparametric Estimation of Functions of Several Variables

A partial spline model is a model for a response as a function of several variables, which is the sum of a smooth function of several variables and a parametric function of the same plus possibly some other variables. Partial spline models in one and several variables, with direct and indirect data, with Gaussian errors and as an extension of GLIM to partially penalized GLIM models are described. Application to the modeling of change of regime in several variables is described. Interaction splines are introduced and described and their potential use for modeling non-linear interactions between variables by semiparametric methods is noted. Reference is made to recent work in efficient computational methods.

Wahba, Grace↗
DOE OSTI2025

Persistent Classification: Understanding Adversarial Attacks by Studying Decision Boundary Dynamics

ABSTRACT There are a number of hypotheses underlying the existence of adversarial examples for classification problems. These include the high‐dimensionality of the data, the high codimension in the ambient space of the data manifolds of interest, and that the structure of machine learning models may encourage classifiers to develop decision boundaries close to data points. This article proposes a new framework for studying adversarial examples that does not depend directly on the distance to the decision boundary. Similarly to the smoothed classifier literature, we define a (natural or adversarial) data point to be ( γ , σ)‐stable if the probability of the same classification is at least for points sampled in a Gaussian neighborhood of the point with a given standard deviation . We focus on studying the differences between persistence metrics along interpolants of natural and adversarial points. We show that adversarial examples have significantly lower persistence than natural examples for large neural networks in the context of the MNIST and ImageNet datasets. We connect this lack of persistence with decision boundary geometry by measuring angles of interpolants with respect to decision boundaries. Finally, we connect this approach with robustness by developing a manifold alignment gradient metric and demonstrating the increase in robustness that can be achieved when training with the addition of this metric.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
NASA NTRS1987

The topology of large-scale structure. I - Topology and the random phase hypothesis

Many models for the formation of galaxies and large-scale structure assume a spectrum of random phase (Gaussian), small-amplitude density fluctuations as initial conditions. In such scenarios, the topology of the galaxy distribution on large scales relates directly to the topology of the initial density fluctuations. Here a quantitative measure of topology - the genus of contours in a smoothed density distribution - is described and applied to numerical simulations of galaxy clustering, to a variety of three-dimensional toy models, and to a volume-limited sample of the CfA redshift survey. For random phase distributions the genus of density contours exhibits a universal dependence on threshold density. The clustering simulations show that a smoothing length of 2-3 times the mass correlation length is sufficient to recover the topology of the initial fluctuations from the evolved galaxy distribution. Cold dark matter and white noise models retain a random phase topology at shorter smoothing lengths, but massive neutrino models develop a cellular topology.

Weinberg, David H.↗
NASA NTRS1988

International Conference on Advances in Communication and Control Systems, 1st, Washington, DC, June 18-20, 1987, Proceedings

Theoretical models of communication and control systems are discussed in reviews and reports. Topics addressed include smoothing and identification for random fields, the information and coding capacities of mismatched Gaussian channels, recursive least-squares estimation and Kalman filtering by systolic arrays, Kemp echo digital filters, a periodic test-scheduling scheme for communication and queuing processes, and receivers for direct-sequence systems. Consideration is given to a distributed-parameter model for detecting cracks in rotors, active control of aeroelastic systems governed by functional differential equations, robust multivariable control of large space structures, finite-rank relatively bounded perturbations of semigroup generators, and sensitivity analysis of convex optimal-control problems.

Declaris, Nicholas↗
NASA NTRS1989

Application of super smooth optics to extra-solar planet detection

The observational constraints on direct detection of extrasolar planets around nearby stars are defined. The design of an instrument capable of meeting these goals requires the efficient reduction of diffracted and scattered light by about three orders of magnitude below the airy wings of the unapodized aperture. The diffraction reduction is achieved through the high-efficiency coronagraph, which uses Gaussian occulting masks in the first focus to concentrate better diffracted light in the pupil plane. Laboratory experiments confirm this high efficiency to the limits set by the experimental setup. Scattering due to midspatial scale figure error in the mirrors must also be reduced to a comparable level.

Terrile, Richard J.↗
DOE OSTI2024

Towards optimal sensor placement for inverse problems in spaces of measures

The objective of this work is to quantify the reconstruction error in sparse inverse problems with measures and stochastic noise, motivated by optimal sensor placement. To be useful in this context, the error quantities must be explicit in the sensor configuration and robust with respect to the source, yet relatively easy to compute in practice, compared to a direct evaluation of the error by a large number of samples. In particular, we consider the identification of a measure consisting of an unknown linear combination of point sources from a finite number of measurements contaminated by Gaussian noise. The statistical framework for recovery relies on two main ingredients: first, a convex but non-smooth variational Tikhonov point estimator over the space of Radon measures and, second, a suitable mean-squared error based on its Hellinger–Kantorovich distance to the ground truth. To quantify the error, we employ a non-degenerate source condition as well as careful linearization arguments to derive a computable upper bound. This leads to asymptotically sharp error estimates in expectation that are explicit in the sensor configuration. Thus they can be used to estimate the expected reconstruction error for a given sensor configuration and guide the placement of sensors in sparse inverse problems.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1990

High spatial resolution 100 micron observations of the M83 bar

A program of high spatial resolution far-infrared observations of galaxies using the Kuiper Airborne Observatory (KAO), was conducted to better understand the role of star formation, the general interstellar radiation field, and non-thermal activity in powering the prodigious far-infrared luminosities seen in spiral and interacting galaxies. Here, researchers present observations of the central region of the well-known barred spiral M83 (NGC 5236). The resultant channel 3 scans for M83 and IRC + 10216, after co-addition and smoothing, are shown. These data show that M83 is extended at 100 microns compared to a point source. A simple Gaussian deconvolution of the M83 data with the point source profile from IRC+10216 gives a full width half maximum (FWHM) of about 19 seconds for M83. By comparison with IRC+10216, researchers obtain a flux for the unresolved component in M83 of about 110 Jy. This is about 1/6 the total flux for M83 (Rice et al. 1988) and about 1/2 the PSC flux. The M83 and IRC+10216 profiles in the cross-scan direction (SE-NW) were also compared, and show that M83 is extended in this direction as well, with a width of about 18 seconds. A comparison of the different channel profiles for M83 and IRC+10216 shows that there is an asymmetry in the M83 data, in that the maximum in the profiles shifts from southeast to northwest as channel number increases. This corresponds to the extension in the bar seen in the CO data. Thus the far-infrared emission in the central region of M83 tends to trace the CO bar. The new 100 micron data is also compared with previous H alpha observations from the literature, to determine how well the far-infrared traces the stellar structure, the star formation as measured by H alpha, and the optical colors.

Smith, B. J.↗