Search NASA⌕ Search

SEARCH · Search NASA

Results for “stochastic gradient”

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 163 records · Page 9

AADL: Anderson Accelerated Deep Learning

We propose a stable, distributed approach to perform AA that accelerates the convergence rate of stochastic first-order optimizers to train neural networks. Differently from previous works, we do not alter neither the scheme to perform AA nor the loss function minimized during the training. To improve robustness against stagnation, we customize general guidelines that suggest to relax the frequency of AA corrections by performing AA only at the end of an entire training epoch. To improve robustness of AA against the stochastic oscillations of first-order optimizers, we average the gradients computed on consecutive stochastic optimization updates. The improved regularity of the converging sequence and the reduced amplitude of stochastic oscillations across consecutive optimization steps allows AA to efficiently extrapolate an improved converging sequence, thereby overcoming limitations of existing approaches to perform AA on stochastic optimization.

Lupo Pasini, Massimiliano [Oak Ridge National Lab.↗

An iterative stochastic ensemble method for parameter estimation of subsurface flow models

Parameter estimation for subsurface flow models is an essential step for maximizing the value of numerical simulations for future prediction and the development of effective control strategies. We propose the iterative stochastic ensemble method (ISEM) as a general method for parameter estimation based on stochastic estimation of gradients using an ensemble of directional derivatives. ISEM eliminates the need for adjoint coding and deals with the numerical simulator as a blackbox. The proposed method employs directional derivatives within a Gauss–Newton iteration. The update equation in ISEM resembles the update step in ensemble Kalman filter, however the inverse of the output covariance matrix in ISEM is regularized using standard truncated singular value decomposition or Tikhonov regularization. We also investigate the performance of a set of shrinkage based covariance estimators within ISEM. The proposed method is successfully applied on several nonlinear parameter estimation problems for subsurface flow models. The efficiency of the proposed algorithm is demonstrated by the small size of utilized ensembles and in terms of error convergence rates.

97 MATHEMATICS AND COMPUTING↗

CO 2 water-alternating-gas injection for enhanced oil recovery: Optimal well controls and half-cycle lengths

We report that CO 2 water-alternating-gas (WAG) injection is an enhanced oil recovery method designed to improve sweep efficiency during CO 2 injection with the injected water to control the mobility of CO 2 and to stabilize the gas front. Optimization of CO 2 -WAG injection is widely regarded as a viable technique for controlling the CO 2 and oil miscible process. Poor recovery from CO 2 -WAG injection can be caused by inappropriately designed WAG parameters. In previous study (Chen and Reynolds, 2016), we proposed an algorithm to optimize the well controls which maximize the life-cycle net-present-value (NPV). However, the effect of injection half-cycle lengths for each injector on oil recovery or NPV has not been well investigated. In this paper, an optimization framework based on augmented Lagrangian method and the newly developed stochastic-simplex-approximate-gradient (StoSAG) algorithm is proposed to explore the possibility of simultaneous optimization of the WAG half-cycle lengths together with the well controls. Finally, the proposed framework is demonstrated with three reservoir examples.

42 ENGINEERING↗

DPM: A deep learning PDE augmentation method with application to large-eddy simulation

A framework is introduced that leverages known physics to reduce overfitting in machine learning for scientific applications. The partial differential equation (PDE) that expresses the physics is augmented with a neural network that uses available data to learn a description of the corresponding unknown or unrepresented physics. Training within this combined system corrects for missing, unknown, or erroneously represented physics, including discretization errors associated with the PDE's numerical solution. For optimization of the network within the PDE, an adjoint PDE is solved to provide high-dimensional gradients, and a stochastic adjoint method (SAM) further accelerates training. Additionally, the approach is demonstrated for large-eddy simulation (LES) of turbulence. High-fidelity direct numerical simulations (DNS) of decaying isotropic turbulence provide the training data used to learn sub-filter-scale closures for the filtered Navier–Stokes equations. Out-of-sample comparisons show that the deep learning PDE method outperforms widely-used models, even for filter sizes so large that they become qualitatively incorrect. It also significantly outperforms the same neural network when a priori trained based on simple data mismatch, not accounting for the full PDE. Measures of discretization errors, which are well-known to be consequential in LES, point to the importance of the unified training formulation's design, which without modification corrects for them. For comparable accuracy, simulation runtime is significantly reduced. A relaxation of the typical discrete enforcement of the divergence-free constraint in the solver is also successful, instead allowing the DPM to approximately enforce incompressibility physics. Since the training loss function is not restricted to correspond directly to the closure to be learned, training can incorporate diverse data, including experimental data.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Capacity assessment and co-optimization of CO 2 storage and enhanced oil recovery in residual oil zones

Residual oil zones (ROZs) are increasingly being commercially exploited using CO 2 -enhanced oil recovery (CO 2 -EOR) method. In this study, CO 2 storage potential, long-term CO 2 fate and oil recovery potential in ROZs are characterized based on a reservoir model for Goldsmith-Landreth San Andres Unit in the Permian Basin. The effects of CO 2 injection rates, well patterns (five-spot and line-drive), well spacings, injection modes (continuous CO 2 injection and water-alternatinggas injection) on the CO 2 retention in the reservoir and the oil production are investigated. After the preliminary assessment of CO 2 storage and EOR potentials in ROZs, we next develop a novel approach based on a newly developed optimization algorithm-Stochastic Simplex Approximate Gradient (StoSAG) and predictive empirical models constructed using machine learning technique to co-optimize CO 2 storage and oil recovery in ROZs. The performance of co-optimization of CO 2 storage and oil recovery is compared with the performance of optimization of only CO 2 storage.

58 GEOSCIENCES↗

Programming Probabilistic Structural Analysis for Parallel Processing Computer

The ultimate goal of this research program is to make Probabilistic Structural Analysis (PSA) computationally efficient and hence practical for the design environment by achieving large scale parallelism. The paper identifies the multiple levels of parallelism in PSA, identifies methodologies for exploiting this parallelism, describes the development of a parallel stochastic finite element code, and presents results of two example applications. It is demonstrated that speeds within five percent of those theoretically possible can be achieved. A special-purpose numerical technique, the stochastic preconditioned conjugate gradient method, is also presented and demonstrated to be extremely efficient for certain classes of PSA problems.

Sues, Robert H.↗

Final Technical Report: Quantification of Uncertainty in Extreme Scale Computations (QUEST)

QUEST is a SciDAC Institute comprising Sandia National Laboratories, Los Alamos National Laboratory, University of Southern California, Massachusetts Institute of Technology, University of Texas at Austin, and Duke University. The mission of QUEST is to: (1) develop a broad class of uncertainty quantification (UQ) methods/tools, and (2) provide UQ expertise and software to other SciDAC projects, thereby enabling/guiding their UQ activities. The Duke effort focused on the development of algorithms and utility software for non-intrusive sparse UQ representations, and on participation in the organization of annual workshops and tutorials to disseminate UQ tools to the community, and to gather input in order to adapt approaches to the needs of SciDAC customers. In particular, fundamental developments were made in (a) multiscale stochastic preconditioners, (b) gradient-based approaches to inverse problems, (c) adaptive pseudo-spectral approximations, (d) stochastic limit cycles, and (e) sensitivity analysis tools for noisy systems. In addition, large-scale demonstrations were performed, namely in the context of ocean general circulation models.

97 MATHEMATICS AND COMPUTING↗

Large deviations in stochastic heat-conduction processes provide a gradient-flow structure for heat conduction

We consider three one-dimensional continuous-time Markov processes on a lattice, each of which models the conduction of heat: the family of Brownian Energy Processes with parameter m (BEP(m)), a Generalized Brownian Energy Process, and the Kipnis-Marchioro-Presutti (KMP) process. The hydrodynamic limit of each of these three processes is a parabolic equation, the linear heat equation in the case of the BEP(m) and the KMP, and a nonlinear heat equation for the Generalized Brownian Energy Process with parameter a (GBEP(a)). We prove the hydrodynamic limit rigorously for the BEP(m), and give a formal derivation for the GBEP(a). We then formally derive the pathwise large-deviation rate functional for the empirical measure of the three processes. These rate functionals imply gradient-flow structures for the limiting linear and nonlinear heat equations. We contrast these gradient-flow structures with those for processes describing the diffusion of mass, most importantly the class of Wasserstein gradient-flow systems. The linear and nonlinear heat-equation gradient-flow structures are each driven by entropy terms of the form -log ρ; they involve dissipation or mobility terms of order ρ² for the linear heat equation, and a nonlinear function of ρ for the nonlinear heat equation.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

An adaptive hierarchical sparse grid collocation algorithm for the solution of stochastic differential equations

In recent years, there has been a growing interest in analyzing and quantifying the effects of random inputs in the solution of ordinary/partial differential equations. To this end, the spectral stochastic finite element method (SSFEM) is the most popular method due to its fast convergence rate. Recently, the stochastic sparse grid collocation method has emerged as an attractive alternative to SSFEM. It approximates the solution in the stochastic space using Lagrange polynomial interpolation. The collocation method requires only repetitive calls to an existing deterministic solver, similar to the Monte Carlo method. However, both the SSFEM and current sparse grid collocation methods utilize global polynomials in the stochastic space. Thus when there are steep gradients or finite discontinuities in the stochastic space, these methods converge very slowly or even fail to converge. In this paper, we develop an adaptive sparse grid collocation strategy using piecewise multi-linear hierarchical basis functions. Hierarchical surplus is used as an error indicator to automatically detect the discontinuity region in the stochastic space and adaptively refine the collocation points in this region. Numerical examples, especially for problems related to long-term integration and stochastic discontinuity, are presented. Comparisons with Monte Carlo and multi-element based random domain decomposition methods are also given to show the efficiency and accuracy of the proposed method.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Stochastic and Deterministic Assembly Processes in Subsurface Microbial Communities

A major goal of microbial community ecology is to understand the forces that structure community composition. Deterministic selection by specific environmental factors is sometimes important, but in other cases stochastic or ecologically neutral processes dominate. Lacking is a unified conceptual framework aiming to understand why deterministic processes dominate in some contexts but not others. Here we work towards such a framework. By testing predictions derived from general ecological theory we aim to uncover factors that govern the relative influences of deterministic and stochastic processes. We couple spatiotemporal data on subsurface microbial communities and environmental parameters with metrics and null models of within and between community phylogenetic composition. Testing for phylogenetic signal in organismal niches showed that more closely related taxa have more similar habitat associations. Community phylogenetic analyses further showed that ecologically similar taxa coexist to a greater degree than expected by chance. Environmental filtering thus deterministically governs subsurface microbial community composition. More importantly, the influence of deterministic environmental filtering relative to stochastic factors was maximized at both ends of an environmental variation gradient. A stronger role of stochastic factors was, however, supported through analyses of phylogenetic temporal turnover. While phylogenetic turnover was on average faster than expected, most pairwise comparisons were not themselves significantly non-random. The relative influence of deterministic environmental filtering over community dynamics was elevated, however, in the most temporally and spatially variable environments. Our results point to general rules governing the relative influences of stochastic and deterministic processes across micro- and macro-organisms.

community assembly↗

A dynamically adaptive wavelet approach to stochastic computations based on polynomial chaos - capturing all scales of random modes on independent grids

Highlights: {yields} New approach for stochastic computations based on polynomial chaos. {yields} Development of dynamically adaptive wavelet multiscale solver using space refinement. {yields} Accurate capture of steep gradients and multiscale features in stochastic problems. {yields} All scales of each random mode are captured on independent grids. {yields} Numerical examples demonstrate the need for different space resolutions per mode. - Abstract: In stochastic computations, or uncertainty quantification methods, the spectral approach based on the polynomial chaos expansion in random space leads to a coupled system of deterministic equations for the coefficients of the expansion. The size of this system increases drastically when the number of independent random variables and/or order of polynomial chaos expansions increases. This is invariably the case for large scale simulations and/or problems involving steep gradients and other multiscale features; such features are variously reflected on each solution component or random/uncertainty mode requiring the development of adaptive methods for their accurate resolution. In this paper we propose a new approach for treating such problems based on a dynamically adaptive wavelet methodology involving space-refinement on physical space that allows all scales of each solution component to be refined independently of the rest. We exemplify this using the convection-diffusion model with random input data and present three numerical examples demonstrating the salient features of the proposed method. Thus we establish a new, elegant and flexible approach for stochastic problems with steep gradients and multiscale features based on polynomial chaos expansions.

97 MATHEMATICS AND COMPUTING↗

An adaptive stochastic sequential quadratic programming with differentiable exact augmented lagrangians

In this study, we consider solving nonlinear optimization problems with a stochastic objective and deterministic equality constraints. We assume for the objective that its evaluation, gradient, and Hessian are inaccessible, while one can compute their stochastic estimates by, for example, subsampling. We propose a stochastic algorithm based on sequential quadratic programming (SQP) that uses a differentiable exact augmented Lagrangian as the merit function. To motivate our algorithm design, we first revisit and simplify an old SQP method Lucidi developed for solving deterministic problems, which serves as the skeleton of our stochastic algorithm. Based on the simplified deterministic algorithm, we then propose a non-adaptive SQP for dealing with stochastic objective, where the gradient and Hessian are replaced by stochastic estimates but the stepsizes are deterministic and prespecified. Finally, we incorporate a recent stochastic line search procedure Paquette and Scheinberg into the non-adaptive stochastic SQP to adaptively select the random stepsizes, which leads to an adaptive stochastic SQP. The global "almost sure" convergence for both non-adaptive and adaptive SQP methods is established. Numerical experiments on nonlinear problems in CUTEst test set demonstrate the superiority of the adaptive algorithm.

97 MATHEMATICS AND COMPUTING↗

Oxygen Ion Heat Rate within Alfvenic Turbulence in the Cusp

The role that the cleft/cusp has in ionosphere-magnetosphere coupling makes it a dynamic and important region. It is directly exposed to the solar wind, making it possible for the entry of electromagnetic energy and precipitating electrons and ions from dayside reconnection and other dayside events. It is also a significant source of ionospheric plasma, contributing largely to the mass loading of the magnetosphere with large fluxes of outflowing ions. Crossing the cusp/cleft near 5100 km, the Polar instruments observe the common correlation of downward Poynting flux, ion energization, soft electron precipitation, broadband extremely low-frequency (BB-ELF) emissions, and density depletions. The dominant power in the BB-ELF emissions is now identified to be from spatially broad, low frequency Alfv nic structures. For a cusp crossing, we determine using the Electric Field Investigation (EFI), that the electric and magnetic field fluctuations are Alfv nic and the electric field gradients satisfy the inequality for stochastic acceleration. With all the Polar 1996 horizontal crossings of the cusp, we determine the O+ heating rate using the Thermal Ion Dynamics Experiment (TIDE) and Plasma Wave Investigation (PWI). We then compare this heating rate to other heating rates assuming the electric field gradient criteria exceeds the limit for stochastic acceleration for the remaining crossings. The comparison suggests that a stochastic acceleration mechanism is operational and the heating is controlled by the transverse spatial scale of the Alfvenic waves.

Coffey, Victoria N.↗

Quantum optimization algorithms: Energetic implications

Since the dawn of quantum computing (QC), theoretical developments like Shor's algorithm proved the conceptual superiority of QC over traditional computing. However, such quantum supremacy claims are difficult to achieve in practice because of the technical challenges of realizing noiseless qubits. In the near future, QC applications will need to rely on noisy quantum devices that offload part of their work to classical devices. One way to achieve this is by using parameterized quantum circuits in optimization or even in machine learning tasks. The energy requirements of quantum algorithms have not yet been studied extensively. Here in this article, we explore several optimization algorithms using both theoretical insights and numerical experiments to understand their impact on energy consumption. Specifically, we highlight why and how algorithms like quantum natural gradient descent, simultaneous perturbation stochastic approximations or circuit learning methods, are at least 2x to 4x more energy efficient than their classical counterparts; why feedback-based quantum optimization is energy-inefficient; and how techniques like Rosalin can improve the energy efficiency of other algorithms by a factor of ≥2 0 x. Finally, we use the NchooseK high-level programming model to run optimization problems on both gate-based quantum computers and quantum annealers. Empirical data indicate that these optimization problems run faster, have better success rates, and consume less energy on quantum annealers than on their gate-based counterparts.

97 MATHEMATICS AND COMPUTING↗

Influence of grain charge gradients on the dynamics of macroparticles in an electrostatic trap

An analytical model of anomalous heating of charged dust grains (macroparticles) caused by their stochastic motion in a bounded plasma volume is proposed. Analytical expressions allowing one to describe the pumping (heating) of interacting grains with additional stochastic energy due to grain charge gradients are derived. The analytical results are verified by numerical simulation of the problem. It is shown that spatial variations in the charges of dust grains can lead to their anomalous heating in laboratory plasma.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Random coordinate descent: A simple alternative for optimizing parameterized quantum circuits

Variational quantum algorithms rely on the optimization of parameterized quantum circuits in noisy settings. The commonly used back-propagation procedure in classical machine learning is not directly applicable in this setting due to the collapse of quantum states after measurements. Thus, gradient estimations constitute a significant overhead in a gradient-based optimization of such quantum circuits. This paper introduces a random coordinate descent algorithm as a practical and easy-to-implement alternative to the full gradient descent algorithm. This algorithm only requires one partial derivative at each iteration. Motivated by the behavior of measurement noise in the practical optimization of parameterized quantum circuits, this paper presents an optimization problem setting that is amenable to analysis. Under this setting, the random coordinate descent algorithm exhibits the same level of stochastic stability as the full gradient approach, making it as resilient to noise. The complexity of the random coordinate descent method is generally no worse than that of the gradient descent and can be much better for various quantum optimization problems with anisotropic Lipschitz constants. Theoretical analysis and extensive numerical experiments validate our findings. Published by the American Physical Society 2024

Ding, Zhiyan (ORCID:000000018863403X)↗

Differential Equation Approximation Using Gradient-Boosted Quantile Regression

The operation of cyber-physical-human (CPH) systems is subject to various epistemic and aleatory uncertainties. Overall trustworthiness of CPH systems relies on the trustworthiness of its components and their interactions. It is important that computational models comprising the cyber component of CPH provide predictions accompanied by a measure of confidence in model outcomes. Uncertainty quantification (UQ) and propagation are especially important in safety critical CPH systems. Gradient-boosted trees is a modeling approach capable both of learning the dynamics of a system and performing UQ. In this paper, we devise a method for using gradient boosting to learn the dynamics of a second order differential equation and estimate uncertainty at the same time. We do this by creating a custom loss function that trains the model to approximate the second derivative of a noisy time series, and to penalize based on a parameter that corresponds to the desired quantile. The resulting gradient boosting model can simulate stochastic trajectories of the system given a single starting point, that is, it can estimate both the expected trajectory and its uncertainty. We show that the uncertainty estimation is well calibrated and that the model can learn the dynamics even in the presence of noise. We demonstrate the approach on a simple cartpole system.

Autonomous systems↗

Solving high-dimensional partial differential equations using deep learning

Developing algorithms for solving high-dimensional partial differential equations (PDEs) has been an exceedingly difficult task for a long time, due to the notoriously difficult problem known as the “curse of dimensionality.” This paper introduces a deep learning-based approach that can handle general high-dimensional parabolic PDEs. To this end, the PDEs are reformulated using backward stochastic differential equations and the gradient of the unknown solution is approximated by neural networks, very much in the spirit of deep reinforcement learning with the gradient acting as the policy function. Numerical results on examples including the nonlinear Black–Scholes equation, the Hamilton–Jacobi–Bellman equation, and the Allen–Cahn equation suggest that the proposed algorithm is quite effective in high dimensions, in terms of both accuracy and cost. Furthermore, this opens up possibilities in economics, finance, operational research, and physics, by considering all participating agents, assets, resources, or particles together at the same time, instead of making ad hoc assumptions on their interrelationships.

97 MATHEMATICS AND COMPUTING↗