Search NASA⌕ Search

SEARCH · Search NASA

Results for “Randomized 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 1,351 records · Page 75

A heterogeneous stochastic FEM framework for elliptic PDEs

We introduce a new concept of sparsity for the stochastic elliptic operator −div(a(x,ω)∇(⋅)), which reflects the compactness of its inverse operator in the stochastic direction and allows for spatially heterogeneous stochastic structure. This new concept of sparsity motivates a heterogeneous stochastic finite element method (HSFEM) framework for linear elliptic equations, which discretizes the equations using the heterogeneous coupling of spatial basis with local stochastic basis to exploit the local stochastic structure of the solution space. We also provide a sampling method to construct the local stochastic basis for this framework using the randomized range finding techniques. The resulting HSFEM involves two stages and suits the multi-query setting: in the offline stage, the local stochastic structure of the solution space is identified; in the online stage, the equation can be efficiently solved for multiple forcing functions. An online error estimation and correction procedure through Monte Carlo sampling is given. Numerical results for several problems with high dimensional stochastic input are presented to demonstrate the efficiency of the HSFEM in the online stage.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Bayesian Monte Carlo Method for Nuclear Data Evaluation

A Bayesian Monte Carlo method is outlined which allows a systematic evaluation of nuclear reactions using TALYS. The result will be either an EXFOR-weighted covariance matrix or a collection of random files, each accompanied by an experiment based weight.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Land surface model parameter optimisation using in situ flux data: comparison of gradient-based versus random search algorithms (a case study using ORCHIDEE v1.9.5.2)

Land surface models (LSMs), which form the land component of earth system models, rely on numerous processes for describing carbon, water and energy budgets, often associated with highly uncertain parameters. Data assimilation (DA) is a useful approach for optimising the most critical parameters in order to improve model accuracy and refine future climate predictions. In this study, we compare two different DA methods for optimising the parameters of seven plant functional types (PFTs) of the ORCHIDEE LSM using daily averaged eddy-covariance observations of net ecosystem exchange and latent heat flux at 78 sites across the globe. We perform a technical investigation of two classes of minimisation methods – local gradient-based (the L-BFGS-B algorithm, limited memory Broyden–Fletcher–Goldfarb–Shanno algorithm with bound constraints) and global random search (the genetic algorithm) – by evaluating their relative performance in terms of the model–data fit and the difference in retrieved parameter values. We examine the performance of each method for two cases: when optimising parameters at each site independently (“single-site” approach) and when simultaneously optimising the model at all sites for a given PFT using a common set of parameters (“multi-site” approach). We find that for the single site case the random search algorithm results in lower values of the cost function (i.e. lower model–data root mean square differences) than the gradient-based method; the difference between the two methods is smaller for the multi-site optimisation due to a smoothing of the cost function shape with a greater number of observations. The spread of the cost function, when performing the same tests with 16 random first-guess parameters, is much larger with the gradient-based method, due to the higher likelihood of being trapped in local minima. When using pseudo-observation tests, the genetic algorithm results in a closer approximation of the true posterior parameter value in the L-BFGS-B algorithm. We demonstrate the advantages and challenges of different DA techniques and provide some advice on using it for the LSM parameter optimisation.

58 GEOSCIENCES↗

Connecting the dots: Semi-analytical and random walk numerical solutions of the diffusion–reaction equation with stochastic initial conditions

We study a system with bimolecular irreversible kinetic reaction A+B→∅ where the underlying transport of reactants is governed by diffusion, and the local reaction term is given by the law of mass action. We consider the case where the initial concentrations are given in terms of an average and a white noise perturbation. Our goal is to solve the diffusion–reaction equation which governs the system, and we tackle it with both analytical and numerical approaches. To obtain an analytical solution, we develop the equations of moments and solve them approximately. To obtain a numerical solution, we develop a grid-less Monte Carlo particle tracking approach, where diffusion is modeled by a random walk of the particles, and reaction is modeled by annihilation of particles. The probability of annihilation is derived analytically from the particles' co-location probability. We rigorously derive the relationship between the initial number of particles in the system and the amplitude of white noise represented by that number. This enables us to compare the particle simulations and the approximate analytical solution and offer an explanation of the late time discrepancies. - Graphical abstract:.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Formation and structural phase transition in Co atomic chains on a Cu(775) surface

The formation of Co atomic chains on a Cu(775) surface is investigated by the kinetic Monte Carlo method. It is found that the length of Co atomic chains formed as a result of self-organization during epitaxial growth is a random quantity and its mean value depends on the parameters of the experiment. The existence of two structural phases in atomic chains is detected using the density functional theory. In the first phase, the separations between an atom and its two nearest neighbors in a chain are 0.230 and 0.280 nm. In the second phase, an atomic chain has identical atomic spacings of 0.255 nm. It is shown that the temperature of the structural phase transition depends on the length of the atomic chain.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

SU-C-BRE-05: PTV Margin Determination Based On Tumor Radiobiological Characteristics and Geometric Uncertainties Derived From Daily Cone- Beam CT Images

Purpose: To determine required PTV margins for ≤1% loss in mean population TCP using systematic (Σ) and random (σ) errors calculated from daily cone-beam CT (CBCT) images of head and neck patients. Methods: Daily CBCT images were acquired for 50 head and neck patients. The CBCT image sets acquired at each fraction were registered with planning CT to obtain positional errors for each patient for each fraction. Systematic and random errors were calculated from data collected for 50 patients as described in IPEM On Target report. CTV delineation uncertainty of 2mm is added quadratically to systematic error. Assuming a spherical target volume, the dose in each voxel of target volume is summed for each fraction in the treatment by shifting the dose grid to calculate mean population TCP inclusive of geometric uncertainties using a Monte Carlo method. These simulations were repeated for the set of Σ and σ in each axis for different PTV margins and drop in TCP for each margin are obtained. In order to study the effect of dose-response curve on PTV margins, two different σα of 0.048 Gy-1 and 0.218 Gy-1 representing steep and shallow dose-response curves are studied. Σ were 2.5, 2.5, 2.1 mm and σ were 0.3, 0.3 0.2 mm respectively in x, y and z axis respectively. Results: PTV margins based on tumor radiobiological characteristics are 4.8, 4.8 and 4 mm in x, y and z axis assuming 25 treatment fractions for σα 0.048 Gy-1 (steep) and 4.2,4.2 and 2.2 for σα of 0.218 Gy-1 (shallow). While the TCP-based margins did not differ much in x and y axis, it is considerably smaller in z axis for shallow DRC. Conclusion: TCP based margins are substantially smaller than physical dose-based margin recipes. This study also demonstrates the importance of considering tumor radiobiological characteristics while deriving margins.

60 APPLIED LIFE SCIENCES↗

The Randomized Freeway Stent Study: Drug-Eluting Balloons Outperform Standard Balloon Angioplasty for Postdilatation of Nitinol Stents in the SFA and PI Segment

Purpose: The prospective randomized multicenter Freeway study evaluated the possible hemodynamic and clinical benefits of primary stent insertion followed by percutaneous transluminal angioplasty (PTA) with drug-eluting balloons (DEB) over post-stent insertion PTA with standard balloons in the treatment of symptomatic femoropopliteal arteriosclerotic lesions. Methods: In total, 204 patients in 13 centers in Germany and Austria were enrolled and randomized to primary stenting followed by either FREEWAY™ drug-eluting balloon or standard PTA balloon angioplasty. The primary endpoint was the rate of clinically driven target lesion revascularization (TLR) at 6 months; the secondary endpoints include TLR rate at 12 months and primary patency, shift in Rutherford classification, ankle–brachial index (ABI) and major adverse events (MAE) at 6 and 12 months. Lesion characteristics and vessel patency were analyzed by an independent and blinded corelab. Results: At 6-month and 12-month follow-up, TLR rate was lower in the DEB arm compared to standard PTA but did not reach statistical significance (4.1% vs. 9.0% p = 0.234 and 7.9% vs. 17.7% p = 0.064, respectively). Primary patency was significantly better for patients treated with the DEB at 6 months (90.3% vs. 69.8% p = 0.001) and 12 months (77.4% vs. 61.0% p = 0.027). Improvement in Rutherford classifications was likewise significantly better for patients in the DEB group at 6 (94.9% vs. 84.3% p = 0.027) and 12 months (95.5% vs. 79.9% p = 0.003). The percentage of patients with an improved ABI of 1.0–1.2 was significantly higher in the DEB group compared to the PTA group at 6 months (55.3% vs. 35.3%; p = 0.015) but without significant difference at 12 months (48.2% vs. 32.9%; p = 0.055). At 6 months, rate of major adverse events (MAE) was 1% in both arms, and at 12 months 2.2% for the DEB and 3.8% for the PTA group. Conclusion: The Freeway Stent Study shows that the usage of DEB as a restenosis prophylaxis seems to be safe and feasible. The 12-month follow-up results give a clear sign in favor of the DEB group.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Ab initio approaches to high-entropy alloys: a comparison of CPA, SQS, and supercell methods

In this work, we present a comparative study of different modeling approaches to the electronic properties of the Hf 0.05 Nb 0.05 Ta 0.8 Ti 0.05 Zr 0.05 high-entropy alloy. Common to our modeling is the methodology to compute the one-particle Green’s function in the framework of density functional theory. We demonstrate that the special quasi-random structures modeling and the supercell, i.e., the locally self-consistent multiple-scattering methods, provide very similar results for the ground state properties such as the spectral function (density of states) and the equilibrium lattice parameter. To reconcile the multiple-scattering single-site coherent potential approximation with the real space supercell methods, we included the effect of screening of the net charges of the alloy components. Based on the analysis of the total energy and spectral functions computed within the density functional theory, we found no signature for the long-range or local magnetic moments formation in the Hf 0.05 Nb 0.05 Ta 0.8 Ti 0.05 Zr 0.05 high-entropy alloy; instead, we find possible superconductivity below ~9K.

36 MATERIALS SCIENCE↗

A simple method for finding the scattering coefficients of quantum graphs

Quantum walks are roughly analogous to classical random walks, and similar to classical walks they have been used to find new (quantum) algorithms. When studying the behavior of large graphs or combinations of graphs, it is useful to find the response of a subgraph to signals of different frequencies. In doing so, we can replace an entire subgraph with a single vertex with variable scattering coefficients. In this paper, a simple technique for quickly finding the scattering coefficients of any discrete-time quantum graph will be presented. These scattering coefficients can be expressed entirely in terms of the characteristic polynomial of the graph’s time step operator. This is a marked improvement over previous techniques which have traditionally required finding eigenstates for a given eigenvalue, which is far more computationally costly. With the scattering coefficients we can easily derive the “impulse response” which is the key to predicting the response of a graph to any signal. This gives us a powerful set of tools for rapidly understanding the behavior of graphs or for reducing a large graph into its constituent subgraphs regardless of how they are connected.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Triangle Method for Dense ReLU Layers [SWR-25-72]

This software is an implementation of the methods for initializing and training neural networks to be more efficient per parameter, described more fully below and in the related publication: In theory, depth should make a ReLU network EXPONENTIALLY more efficient by enabling it to produce an exponential number of piecewise linear sections in its output. This reasoning is largely based on the work of mathematicians that have hand-constructed networks that make good use of depth. In practice however, even very deep ReLU networks that have been randomly initialized will behave identically to their shallow counterparts - missing an entire exponential dimension of efficiency. The triangle method is a first attempt at realizing the exponential potential of deep networks. Instead of randomly setting weights, we force pairs of neurons in each layer learn to build triangles (i.e. functions from [0,1] -> [0,1] that look like triangles). This is a very efficient pattern for generating lots of linear pieces because composing two triangular functions doubles the number of pieces with each composition. The triangle method is more than just a different initialization, it is a new paradigm of training. Instead of making direct updates to the matrix weights, we do an extra step of backpropagation to collect the derivatives of the loss function with respect to the shapes of the triangles, training them to tilt left or right. This process essentially holds the networks hand throughout the loss landscape and forces it to always use depth effectively by producing triangular shapes internally. This can produce several orders of magnitude of improvement on convex one-dimensional regression problems. Much more theoretical work is needed to realize its full potential beyond this context, but the implementation in this repository will still work in arbitrary numbers of dimensions. The file Triangle_Method.py is a generalized form of the method that will build each neuron its own custom 1-d convex activation function (with exponential efficiency). Example usage on one dimensional problems can be found in Example_Usage.ipynb and an example of using this in a real neural network can be found in Example_VGG16_CIFAR10.ipynb.

Milkert, Max [National Renewable Energy Laboratory↗

SU-E-E-05: Improving Contouring Precision and Consistency for Physicians-In-Training with Simple Lab Experiments

Purpose: Target contouring for high-dose treatments such as radiosurgery of brain metastases is highly critical in eliminating marginal failure and reducing complications as shown by recent clinical studies. In order to improve contouring accuracy and practice consistency for the procedure, we introduced a self-assessed physics lab practice for the physicians-in-training. Methods: A set of commercially acquired high-precision PMMA plastic spheres were randomly embedded in a Styrofoam block and then scanned with the CT/MR via the clinical procedural imaging protocol. A group of first-year physicians-in-training (n=6) from either neurosurgery or radiation oncology department were asked to contour the scanned objects (diameter ranged from 0.4 cm to 3.8 cm). These user-defined contours were then compared with the ideal contour sets of object shape for self assessments to determine the maximum areas of the observed discrepancies and method of improvements. Results: The largest discrepancies from initial practice were consistently found to be located near the extreme longitudinal portions of the target for all the residents. Discrepancy was especially prominent when contouring small objects < 1.0 cm in diameters. For example, the mean volumes rendered from the initial contour data set differed from the ideal data set by 7.7%±6.6% for the participants (p> 0.23 suggesting agreement cannot be established). However, when incorporating a secondary imaging scan such as reconstructed coronal or sagittal images in a repeat practice, the agreement was dramatically improved yielding p<0.02 in agreement with the reference data set for all the participants. Conclusion: A simple physics lab revealed a common pitfall in contouring small metastatic brain tumors for radiosurgical procedures and provided a systematic tool for physicians-in-training in improving their clinical contouring skills. Dr Ma is current a board member of international stereotactic radiosurgical society.

60 APPLIED LIFE SCIENCES↗

Learning and discovering multiple solutions using physics-informed neural networks with random initialization and deep ensemble

In this work we explore the capability of physics-informed neural networks (PINNs) to discover multiple solutions. Many real-world phenomena governed by nonlinear differential equations (DEs), such as fluid flow, exhibit multiple solutions under the same conditions, yet capturing this solution multiplicity remains a significant challenge. A key difficulty lies in providing appropriate initial conditions or guesses, as widely used time-marching schemes and Newton’s method are highly sensitive to these choices when solving complex computational problems. While machine learning models, particularly PINNs, have shown promise in solving DEs, their ability to capture multiple solutions remains underexplored. In this work, we propose a simple and practical approach using PINNs to learn and discover multiple solutions. We first demonstrate that PINNs, when combined with random initialization and deep ensemble method—originally developed for uncertainty quantification—can effectively uncover multiple solutions to nonlinear ordinary and partial DEs. Although training large ensembles of PINNs may appear computationally demanding, this can be done efficiently using vectorization techniques supported by modern deep learning frameworks, allowing many networks to be trained simultaneously. Our approach highlights the critical role of initialization in shaping solution diversity, addressing an often-overlooked aspect of machine learning for scientific computing. Furthermore, we propose utilizing PINN-generated solutions as initial conditions or initial guesses for conventional numerical solvers to enhance accuracy and efficiency in capturing multiple solutions. Extensive numerical experiments, including the Allen–Cahn equation and cavity flow, where our approach successfully identifies both stable and unstable solutions, validate the effectiveness of our method. These findings establish a general and efficient framework for addressing solution multiplicity in nonlinear DEs.

97 MATHEMATICS AND COMPUTING↗

Matter-wave localization in a random potential

By numerical and variational solutions of the Gross-Pitaevskii equation, we studied the localization of a noninteracting and weakly interacting Bose-Einstein condensate (BEC) in a disordered cold atom lattice and a speckle potential. In the case of a single BEC fragment, the variational analysis produced good results. For a weakly disordered potential, the localized BEC's are found to have an exponential tail as in the weak Anderson localization. We also investigated the expansion of a noninteracting BEC in these potentials. We find that the BEC will be locked in an appropriate localized state after an initial expansion and will execute breathing oscillation around a mean shape when a BEC at equilibrium in a harmonic trap is suddenly released into a disorder potential.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Invariant functions for the Lyapunov exponents of random matrices

A new approach to the study of Lyapunov exponents of random matrices is presented. We prove that any family of nonnegative (dxd)-matrices has a continuous concave invariant functional on R{sup d}{sub +}. Under some standard assumptions on the matrices, this functional is strictly positive, and the coefficient corresponding to it is equal to the largest Lyapunov exponent. As a corollary we obtain asymptotics for the expected value of the logarithm of norms of matrix products and of their spectral radii. Another corollary gives new upper and lower bounds for the Lyapunov exponent, and an algorithm for computing it for families of nonnegative matrices. We consider possible extensions of our results to general nonnegative matrix families and present several applications and examples. Bibliography: 29 titles.

97 MATHEMATICS AND COMPUTING↗

Measuring representative volume elements from high-resolution grain-scale strain fields

Most crystalline materials present a highly heterogeneous response at the microscale, which can be affected by both internal factors (such as microstructural parameters) and external factors (such as loading). Relating microscale inhomogeneities to the macroscale response of a material requires the use of homogenisation techniques, usually based on the concept of a representative volume element (RVE)—the smallest volume of material that represents the global average response. Here, in this work, we present a new and robust experimental method of measuring the size of a strain-based RVE from high-resolution grain-scale strain fields obtained using digital image correlation (DIC). The proposed method is based on the statistical (stereological) nature of the RVE, which has been widely adopted in numerical studies, and involves dividing a strain field into randomly selected regions of varying sizes and statistically analysing the distributions of average strains within them. To validate the new method, we generate a large number of synthetic strain fields from a fractional Gaussian noise algorithm. The proposed stereological method is shown to be capable of producing reliable RVE measurements from a very large range of possible microscale strain fields while at the same time being robust in that it can produce RVE measurement results even in cases where other existing methods may be unable to do so. The proposed method has a low field-of-view requirement, only needing a field-of-view about 1.2 times as large as the RVE to produce reliable measurements. In addition, the stereological method offers significant flexibility since its statistical nature allows for control over how strict the RVE measurement should be in each case.

36 MATERIALS SCIENCE↗

On Collisions Times of ‘Self-Sorting’ Interacting Particles in One-Dimension with Random Initial Positions and Velocities

We investigate a one-dimensional system of N particles, initially distributed with random positions and velocities, interacting through binary collisions. The collision rule is such that there is a time after which the N particles do not interact and become sorted according to their velocities. When the collisions are elastic, we derive asymptotic distributions for the final collision time of a single particle and the final collision time of the system as the number of particles approaches infinity, under different assumptions for the initial distributions of the particles’ positions and velocities. For comparison, a numerical investigation is carried out to determine how a non-elastic collision rule, which conserves neither momentum nor energy, affects the median collision time of a particle and the median final collision time of the system.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

A weighted ℓ{sub 1}-minimization approach for sparse polynomial chaos expansions

This work proposes a method for sparse polynomial chaos (PC) approximation of high-dimensional stochastic functions based on non-adapted random sampling. We modify the standard ℓ{sub 1}-minimization algorithm, originally proposed in the context of compressive sampling, using a priori information about the decay of the PC coefficients, when available, and refer to the resulting algorithm as weightedℓ{sub 1}-minimization. We provide conditions under which we may guarantee recovery using this weighted scheme. Numerical tests are used to compare the weighted and non-weighted methods for the recovery of solutions to two differential equations with high-dimensional random inputs: a boundary value problem with a random elliptic operator and a 2-D thermally driven cavity flow with random boundary condition.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Localization dynamics in a centrally coupled system

In systems in which interactions couple a central degree of freedom and a bath, one would expect signatures of the bath's phase to be reflected in the dynamics of the central degree of freedom. This has been recently explored in connection with many-body localized baths coupled with a central qubit or a single-cavity mode - systems with growing experimental relevance in various platforms. Such models also have an interesting connection with Floquet many-body localization via quantizing the external drive, although this has been relatively unexplored. Here we adapt the multilayer multiconfigurational time-dependent Hartree (ML-MCTDH) method, a well-known tree tensor network algorithm, to numerically simulate the dynamics of a central degree of freedom, represented by a d-level system (qudit), coupled to a disordered interacting one-dimensional spin bath. ML-MCTDH allows us to reach ≈10 2 lattice sites, a far larger system size than what is feasible with exact diagonalization or kernel polynomial methods. From the intermediate time dynamics, we find a well-defined thermodynamic limit for the qudit dynamics upon appropriate rescaling of the system-bath coupling. The spin system shows similar scaling collapse in the Edward-Anderson spin-glass order parameter or entanglement entropy at relatively short times. At longer timescales, we see slow growth of the entanglement, which may arise from dephasing mechanisms in the localized system or long-range interactions mediated by the central degree of freedom. Similar signs of localization are shown to appear as well with unscaled system-bath coupling.

1-dimensional spin chains↗