Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Regularization”

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 1,387 records · Page 77

Spherical time-encoded radiation imaging simulations

Radiation source localization is important for nuclear nonproliferation and can be obtained using time-encoded imaging systems with unsegmented detectors. A scintillation crystal can be used with a moving coded-aperture mask to vary the detected count rate produced from radiation sources in the far field. The modulation of observed counts over time can be used to reconstruct an image with the known coded-aperture mask pattern. Current time-encoded imaging systems incorporate cylindrical coded-aperture masks and have limits to their fully coded imaging field-of-view. This work focuses on expanding the field-of-view to 4π by using a novel spherical coded-aperture mask. A regular icosahedron is used to approximate a spherical mask. This icosahedron consists of 20 equilateral triangles; the faces of which are each subdivided into four equilateral triangle-shaped voxels which are then projected onto a spherical surface, creating an 80-voxel coded-aperture mask. Furthermore, these polygonal voxels can be made from high-Z materials for gamma-ray modulation and/or low-Z materials for neutron modulation. In this work, we present Monte Carlo N-Particle (MCNP) simulations and simple models programmed in Mathematica to explore image reconstruction capabilities of this 80-voxel coded-aperture mask.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Analytic gradients for the single-reference driven similarity renormalization group second-order perturbation theory

Herein we derive and implement analytic energy gradients for the single-reference driven similarity renormalization group second-order perturbation theory (DSRG-PT2). The resulting equations possess an asymptotic scaling that is identical to that of the second-order Møller–Plesset perturbation theory (MP2), indicating that the exponential regularizer in the DSRG equations does not introduce formal difficulties in the gradient theory. We apply the DSRG-PT2 method to optimizing the geometries of 15 small molecules. The equilibrium bond lengths computed with DSRG-PT2 are found similar to those of MP2, yielding a mean absolute error of 0.0033 Å and a standard deviation of 0.0045 Å when compared with coupled cluster with singles, doubles, and perturbative triples.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An in-silico NMR laboratory for nuclear magnetic shieldings computed via finite fields: Exploring nucleus-specific renormalizations of MP2 and MP3

Here we developed and implemented a method-independent, fully numerical, finite difference approach to calculating nuclear magnetic resonance shieldings, using gauge-including atomic orbitals. The resulting capability can be used to explore non-standard methods, given only the energy as a function of finite-applied magnetic fields and nuclear spins. For example, standard second-order Møller-Plesset theory (MP2) has well-known efficacy for 1H and 13C shieldings and known limitations for other nuclei such as 15N and 17O. It is, therefore, interesting to seek methods that offer good accuracy for 15N and 17O shieldings without greatly increased compute costs, as well as exploring whether such methods can further improve 1H and 13C shieldings. Using a small molecule test set of 28 species, we assessed two alternatives: κ regularized MP2 (κ-MP2), which provides energy-dependent damping of large amplitudes, and MP2.X, which includes a variable fraction, X, of third-order correlation (MP3). The aug-cc-pVTZ basis was used, and coupled cluster with singles and doubles and perturbative triples [CCSD(T)] results were taken as reference values. Our κ-MP2 results reveal significant improvements over MP2 for 13C and 15N, with the optimal κ value being element-specific. κ-MP2 with κ = 2 offers a 30% rms error reduction over MP2. For 15N, κ-MP2 with κ = 1.1 provides a 90% error reduction vs MP2 and a 60% error reduction vs CCSD. On the other hand, MP2.X with a scaling factor of 0.6 outperformed CCSD for all heavy nuclei. These results can be understood as providing renormalization of doubles amplitudes to partially account for neglected triple and higher substitutions and offer promising opportunities for future applications.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

A Consensus Equilibrium Approach for 3-D Land Seismic Shots Recovery

Physical and budget constraints often result in inadequate sampling for accurate subsurface imaging. Preprocessing approaches, such as missing trace interpolation, are typically employed to enhance seismic data in such cases. The compressed sensing (CS) framework has been applied for modeling missing seismic data, which is estimated by sparsity-based computational algorithms. While existing work mainly focuses on recovering missing traces resulting from receiver subsampling, source subsampling has greater economical advantages, as sources are more expensive than receivers. Moreover, stronger image models different from sparsity have not been explored for source recovery. This work presents a consensus equilibrium (CE) approach to recover missing seismic shots, which enables to incorporate various regularization operators modeling different data priors. Here, simulation results from a real 3-D land seismic dataset demonstrate that the CE approach provides more accurate estimations of the linear and hyperbolic events in the recovered shots, compared with pure sparsity-based reconstructions.

58 GEOSCIENCES↗

Distinguishing one from many using super-resolution compressive sensing

We present that distinguishing whether a signal corresponds to a single source or a limited number of highly overlapping point spread functions (PSFs) is a ubiquitous problem across all imaging scales, whether detecting receptor-ligand interactions in cells or detecting binary stars. Super-resolution imaging based upon compressed sensing exploits the relative sparseness of the point sources to successfully resolve sources which may be separated by much less than the Rayleigh criterion. However, as a solution to an underdetermined system of linear equations, compressive sensing requires the imposition of constraints which may not always be valid. One typical constraint is that the PSF is known. However, the PSF of the actual optical system may reflect aberrations not present in the theoretical ideal optical system. Even when the optics are well characterized, the actual PSF may reflect factors such as non-uniform emission of the point source (e.g. fluorophore dipole emission). As such, the actual PSF may differ from the PSF used as a constraint. Similarly, multiple different regularization constraints have been suggested including the l 1 -norm, l 0 -norm, and generalized Gaussian Markov random fields (GGMRFs), each of which imposes a different constraint. Other important factors include the signal-to-noise ratio of the point sources and whether the point sources vary in intensity. In this work, we explore how these factors influence super-resolution image recovery robustness, determining the sensitivity and specificity. In conclusion, we determine an approach that is more robust to the types of PSF errors present in actual optical systems.

42 ENGINEERING↗

The Auxiliary Space Preconditioner for the de Rham Complex

Here we generalize the construction and analysis of auxiliary space preconditioners to the $n$-dimensional finite element subcomplex of the de Rham complex. These preconditioners are based on a generalization of a decomposition of Sobolev space functions into a regular part and a potential. A discrete version is easily established using the tools of finite element exterior calculus. We then discuss the four-dimensional de Rham complex in detail. By identifying forms in four dimensions with simple proxies, form operations are written out in terms of familiar algebraic operations on matrices, vectors, and scalars. This provides the basis for our implementation of the preconditioners in four dimensions. Extensive numerical experiments illustrate their performance, practical scalability, and parameter robustness, all in accordance with the theory.

4D↗

Error Estimates for the Optimal Control of a Parabolic Fractional PDE

In this work, we consider the integral definition of the fractional Laplacian and analyze a linear-quadratic optimal control problem for the so-called fractional heat equation; control constraints are also considered. We derive existence and uniqueness results, first order optimality conditions, and regularity estimates for the optimal variables. To discretize the state equation we propose a fully discrete scheme that relies on an implicit finite difference discretization in time combined with a piecewise linear finite element discretization in space. We derive stability results and a novel $L^2(0,T;L^2(\Omega))$ a priori error estimate. On the basis of the aforementioned solution technique, we propose a fully discrete scheme for our optimal control problem that discretizes the control variable with piecewise constant functions, and we derive a priori error estimates for it. We illustrate the theory with one- and two-dimensional numerical experiments.

97 MATHEMATICS AND COMPUTING↗

On a nonlocal Cahn–Hilliard model permitting sharp interfaces

A nonlocal Cahn–Hilliard model with a non-smooth potential of double-well obstacle type that promotes sharp interfaces in the solution is presented. To capture long-range interactions between particles, a nonlocal Ginzburg–Landau energy functional is defined which recovers the classical (local) model as the extent of nonlocal interactions vanish. In contrast to the local Cahn–Hilliard problem that always leads to diffuse interfaces, the proposed nonlocal model can lead to a strict separation into pure phases of the substance. In this work, the lack of smoothness of the potential is essential to guarantee the aforementioned sharp-interface property. Mathematically, this introduces additional inequality constraints that, in a weak formulation, lead to a coupled system of variational inequalities which at each time instance can be restated as a constrained optimization problem. We prove the well-posedness and regularity of the semi-discrete and continuous in time weak solutions, and derive the conditions under which pure phases are admitted. Moreover, we develop discretizations of the problem based on finite element methods and implicit–explicit time-stepping methods that can be realized efficiently. Finally, we illustrate our theoretical findings through several numerical experiments in one and two spatial dimensions that highlight the differences in features of local and nonlocal solutions and also the sharp interface properties of the nonlocal model.

97 MATHEMATICS AND COMPUTING↗

SINGULARITIES OF GENERALIZED PARTON DISTRIBUTIONS

We discuss recent developments in building models for generalized parton distributions (GPDs) that are based on the formalism of double distributions (DDs). A special attention is given to a careful analysis of the singularity structure of DDs. The DD formalism is applied to construction of a model GPDs with a singular Regge behavior. Within the developed DD-based approach, we discuss the structure of GPD sum rules. It is shown that separation of DDs into the so-called ``plus'' part and the D-term part may be treated as a renormalization procedure for the GPD sum rules. This approach is compared with an alternative prescription based on analytic regularization.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

A nonsmooth nonconvex optimization algorithm for two-stage optimization problems

An optimization algorithm for a group of nonsmooth nonconvex problems inspired by two-stage stochastic programming problems is proposed. The main challenges for these problems include (1) the problems lack the popular lower-type properties such as prox-regularity assumed in many nonsmooth nonconvex optimization algorithms, (2) the objective can not be analytically expressed and (3) the evaluation of function values and subgradients are computationally expensive. To address these challenges, this report first examines the properties that exist in many two-stage problems, specifically upper-C 2 objectives. Then, we show that quadratic penalty method for securityconstrained alternating current optimal power flow (SCACOPF) contingency problems can make the contingency solution functions upper-C 2 . Based on these observations, a simplified bundle algorithm that bears similarity to sequential quadratic programming (SQP) method is proposed. It is more efficient in implementation and computation compared to conventional bundle methods. Global convergence analysis of the algorithm is presented under novel and reasonable assumptions. The proposed algorithm therefore fills the gap of theoretical convergence for smoothed SCACOPF problems. The inconsistency that might arise in our treatment of the constraints are addressed through a penalty algorithm whose convergence analysis is also provided. Finally, theoretical capabilities and numerical performance of the algorithm are demonstrated through numerical examples.

97 MATHEMATICS AND COMPUTING↗

BUTTER - Empirical Deep Learning Dataset

The BUTTER Empirical Deep Learning Dataset represents an empirical study of the deep learning phenomena on dense fully connected networks, scanning across thirteen datasets, eight network shapes, fourteen depths, twenty-three network sizes (number of trainable parameters), four learning rates, six minibatch sizes, four levels of label noise, and fourteen levels of L1 and L2 regularization each. Multiple repetitions (typically 30, sometimes 10) of each combination of hyperparameters were preformed, and statistics including training and test loss (using a 80% / 20% shuffled train-test split) are recorded at the end of each training epoch. In total, this dataset covers 178 thousand distinct hyperparameter settings ("experiments"), 3.55 million individual training runs (an average of 20 repetitions of each experiments), and a total of 13.3 billion training epochs (three thousand epochs were covered by most runs). Accumulating this dataset consumed 5,448.4 CPU core-years, 17.8 GPU-years, and 111.2 node-years.

Array↗

Ps and Qs: Quantization-Aware Pruning for Efficient Low Latency Neural Network Inference

Efficient machine learning implementations optimized for inference in hardware have wide-ranging benefits, depending on the application, from lower inference latency to higher data throughput and reduced energy consumption. Two popular techniques for reducing computation in neural networks are pruning, removing insignificant synapses, and quantization, reducing the precision of the calculations. In this work, we explore the interplay between pruning and quantization during the training of neural networks for ultra low latency applications targeting high energy physics use cases. Techniques developed for this study have potential applications across many other domains. We study various configurations of pruning during quantization-aware training, which we term quantization-aware pruning, and the effect of techniques like regularization, batch normalization, and different pruning schemes on performance, computational complexity, and information content metrics. We find that quantization-aware pruning yields more computationally efficient models than either pruning or quantization alone for our task. Further, quantization-aware pruning typically performs similar to or better in terms of computational efficiency compared to other neural architecture search techniques like Bayesian optimization. Surprisingly, while networks with different training configurations can have similar performance for the benchmark application, the information content in the network can vary significantly, affecting its generalizability.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Equation Discovery Using Fast Function Extraction: a Deterministic Symbolic Regression Approach

Advances in machine learning (ML) coupled with increased computational power have enabled identification of patterns in data extracted from complex systems. ML algorithms are actively being sought in recovering physical models or mathematical equations from data. This is a highly valuable technique where models cannot be built using physical reasoning alone. In this paper, we investigate the application of fast function extraction (FFX), a fast, scalable, deterministic symbolic regression algorithm to recover partial differential equations (PDEs). FFX identifies active bases among a huge set of candidate basis functions and their corresponding coefficients from recorded snapshot data. This approach uses a sparsity-promoting technique from compressive sensing and sparse optimization called pathwise regularized learning to perform feature selection and parameter estimation. Furthermore, it recovers several models of varying complexity (number of basis terms). FFX finally filters out many identified models using non-dominated sorting and forms a Pareto front consisting of optimal models with respect to minimizing complexity and test accuracy. Numerical experiments are carried out to recover several ubiquitous PDEs such as wave and heat equations among linear PDEs and Burgers, Korteweg–de Vries (KdV), and Kawahara equations among higher-order nonlinear PDEs. Additional simulations are conducted on the same PDEs under noisy conditions to test the robustness of the proposed approach.

42 ENGINEERING↗

Material Discrimination for Treaty Verification with Multi-energy, X-ray Radiography

As nuclear warhead stockpiles are reduced under current and future arms treaties, accuracy in accountancy of material stockpiles becomes increasingly important. Image-based active interrogation offers advantages to spectroscopic detection in its ability to measure the location and extent of a material, or materials, of interest from a declaration. However, the detail that imaging provides is often viewed as too intrusive due to its potential to disseminate sensitive information. In this work, we present a method for reducing multi-energy, x-ray radiography data to a few important attributes, based on declarations from behind an information barrier, which can then be used to confirm or deny a declaration. We build on previous work by improving the physics modeling; considering currently attainable, multiple-endpoint x-ray systems; and posing the problem as a nonlinear, inverse problem. Regularization is added to the problem, which smooths the solution and stabilizes an otherwise unstable solution. Here we show the ability to discriminate high-atomic–number materials from others with simulated single-view multiple-endpoint radiography data and present results from initial bench-top measurements.

imaging↗

Anisotropic seismic-waveform inversion: Application to a seismic velocity model from Eleven-Mile Canyon in Nevada

Accurate imaging and characterization of fracture zones is crucial for geothermal energy exploration. Aligned fractures within fracture zones behave as anisotropic media for seismic-wave propagation. The anisotropic properties in fracture zones introduce extra difficulties for seismic imaging and waveform inversion. We have recently developed a new anisotropic elastic-waveform inversion method using a modified total-variation regularization scheme and a wave-energy-base preconditioning technique. Our new inversion method uses the parameterization of elasticity constants to describe anisotropic media, and hence it can properly handle arbitrary anisotropy. We apply our new inversion method to a seismic velocity model along a 2D-line seismic data acquired at Eleven-Mile Canyon located at the Southern Dixie Valley in Nevada for geothermal energy exploration. Our inversion results show that anisotropic elastic-waveform inversion has potential to reconstruct subsurface anisotropic elastic parameters for imaging and characterization of fracture zones.

15 GEOTHERMAL ENERGY↗

Association Between Anticholinergic Medication Use and Cognition, Brain Metabolism, and Brain Atrophy in Cognitively Normal Older Adults

Importance of this Paper: The use of anticholinergic (AC) medication is linked to cognitive impairment and an increased risk of dementia. To our knowledge, this is the first study to investigate the association between AC medication use and neuroimaging biomarkers of brain metabolism and atrophy as a proxy for understanding the underlying biology of the clinical effects of AC medications. Objective: To assess the association between AC medication use and cognition, glucose metabolism, and brain atrophy in cognitively normal older adults from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) and the Indiana Memory and Aging Study (IMAS). Design, Setting, and Participants: The ADNI and IMAS are longitudinal studies with cognitive, neuroimaging, and other data collected at regular intervals in clinical and academic research settings. For the participants in the ADNI, visits are repeated 3, 6, and 12 months after the baseline visit and then annually. For the participants in the IMAS, visits are repeated every 18 months after the baseline visit (402 cognitively normal older adults in the ADNI and 49 cognitively normal older adults in the IMAS were included in the present analysis). Participants were either taking (hereafter referred to as the AC + participants [52 from the ADNI and 8 from the IMAS]) or not taking (hereafter referred to as the AC - participants [350 from the ADNI and 41 from the IMAS]) at least 1 medication with medium or high AC activity. Data analysis for this study was performed in November 2015. Main Outcomes and Measures: Cognitive scores, mean fludeoxyglucose F 18 standardized uptake value ratio (participants from the ADNI only), and brain atrophy measures from structural magnetic resonance imaging were compared between AC + participants and AC - participants after adjusting for potential confounders. The total AC burden score was calculated and was related to target measures. The association of AC use and longitudinal clinical decline (mean [SD] follow-up period, 32.1 [24.7] months [range, 6-108 months]) was examined using Cox regression. Results: The 52 AC + participants (mean [SD] age, 73.3 [6.6] years) from the ADNI showed lower mean scores on Weschler Memory Scale–Revised Logical Memory Immediate Recall (raw mean scores: 13.27 for AC + participants and 14.16 for AC - participants; P = .04) and the Trail Making Test Part B (raw mean scores: 97.85 seconds for AC + participants and 82.61 seconds for AC - participants; P = .04) and a lower executive function composite score (raw mean scores: 0.58 for AC + participants and 0.78 for AC - participants; P = .04) than the 350 AC - participants (mean [SD] age, 73.3 [5.8] years) from the ADNI. Reduced total cortical volume and temporal lobe cortical thickness and greater lateral ventricle and inferior lateral ventricle volumes were seen in the AC + participants relative to the AC - participants. Conclusions and Relevance of this Paper: The use of AC medication was associated with increased brain atrophy and dysfunction and clinical decline. Thus and finally, use of AC medication among older adults should likely be discouraged if alternative therapies are available.

60 APPLIED LIFE SCIENCES↗

Linear functional minimization for inverse modeling

In this paper, we present a novel inverse modeling strategy to estimate spatially distributed parameters of nonlinear models. The maximum a posteriori (MAP) estimators of these parameters are based on a likelihood functional, which contains spatially discrete measurements of the system parameters and spatiotemporally discrete measurements of the transient system states. The piecewise continuity prior for the parameters is expressed via Total Variation (TV) regularization. The MAP estimator is computed by minimizing a nonquadratic objective equipped with the TV operator. We apply this inversion algorithm to estimate hydraulic conductivity of a synthetic confined aquifer from measurements of conductivity and hydraulic head. The synthetic conductivity field is composed of a low-conductivity heterogeneous intrusion into a high-conductivity heterogeneous medium. Our algorithm accurately reconstructs the location, orientation, and extent of the intrusion from the steady-state data only. Finally, addition of transient measurements of hydraulic head improves the parameter estimation, accurately reconstructing the conductivity field in the vicinity of observation locations.

47 OTHER INSTRUMENTATION↗

Global modeling of ULF waves at Mercury

Abstract ULF waves in the ion cyclotron frequency range waves are regularly observed at Mercury's magnetosphere. Although previous statistical studies have shown that ULF waves are primarily compressional near the equator and transverse with linear polarization at higher latitude, the underlying reason for this distribution of wave polarization has not been understood. In order to address this key question, we have developed a two‐dimensional, finite element code that solves the full wave equations in global magnetospheric geometry. Using this code, we show that (1) efficient mode conversion from the fast compressional waves to the ion‐ion hybrid resonance occurs at Mercury consistent with previous calculations; (2) such mode‐converted waves globally oscillate similar to field line resonance at Earth; and (3) compressional wave energy is primarily localized near the equator, while field‐aligned transverse, linearly polarized waves generated by mode conversion at the ion‐ion hybrid resonance radiate to higher latitude. Based on these wave solutions, we suggest that the strong transverse component of observed ULF waves at Mercury in high magnetic latitude can be explained as excitation of the field line resonant waves at the ion‐ion hybrid resonance.

Kim, Eun‐Hwa↗