Search NASA⌕ Search

SEARCH · Search NASA

Results for “randomized algorithm”

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 685 records · Page 38

Phase diagram of the disordered Bose-Hubbard model

We establish the phase diagram of the disordered three-dimensional Bose-Hubbard model at unity filling which has been controversial for many years. The theorem of inclusions, proven by Pollet et al. [Phys. Rev. Lett. 103, 140402 (2009)] states that the Bose-glass phase always intervenes between the Mott insulating and superfluid phases. Here, we note that assumptions on which the theorem is based exclude phase transitions between gapped (Mott insulator) and gapless phases (Bose glass). The apparent paradox is resolved through a unique mechanism: such transitions have to be of the Griffiths type when the vanishing of the gap at the critical point is due to a zero concentration of rare regions where extreme fluctuations of disorder mimic a regular gapless system. An exactly solvable random transverse field Ising model in one dimension is used to illustrate the point. A highly nontrivial overall shape of the phase diagram is revealed with the worm algorithm. The phase diagram features a long superfluid finger at strong disorder and on-site interaction. Moreover, bosonic superfluidity is extremely robust against disorder in a broad range of interaction parameters; it persists in random potentials nearly 50 ({exclamation_point}) times larger than the particle half-bandwidth. Finally, we comment on the feasibility of obtaining this phase diagram in cold-atom experiments, which work with trapped systems at finite temperature.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

SU-F-J-65: Prediction of Patient Setup Errors and Errors in the Calibration Curve from Prompt Gamma Proton Range Measurements

Purpose: To understand the extent to which the prompt gamma camera measurements can be used to predict the residual proton range due to setup errors and errors in the calibration curve. Methods: We generated ten variations on a default calibration curve (CC) and ten corresponding range maps (RM). Starting with the default RM, we chose a square array of N beamlets, which were then rotated by a random angle θ and shifted by a random vector s. We added a 5% distal Gaussian noise to each beamlet in order to introduce discrepancies that exist between the ranges predicted from the prompt gamma measurements and those simulated with Monte Carlo algorithms. For each RM, s, θ, along with an offset u in the CC, were optimized using a simple Euclidian distance between the default ranges and the ranges produced by the given RM. Results: The application of our method lead to the maximal overrange of 2.0mm and underrange of 0.6mm on average. Compared to the situations where s, θ, and u were ignored, these values were larger: 2.1mm and 4.3mm. In order to quantify the need for setup error corrections, we also performed computations in which u was corrected for, but s and θ were not. This yielded: 3.2mm and 3.2mm. The average computation time for 170 beamlets was 65 seconds. Conclusion: These results emphasize the necessity to correct for setup errors and the errors in the calibration curve. The simplicity and speed of our method makes it a good candidate for being implemented as a tool for in-room adaptive therapy. This work also demonstrates that the Prompt gamma range measurements can indeed be useful in the effort to reduce range errors. Given these results, and barring further refinements, this approach is a promising step towards an adaptive proton radiotherapy.

60 APPLIED LIFE SCIENCES↗

i- flow: High-dimensional integration and sampling with normalizing flows

In many fields of science, high-dimensional integration is required. Numerical methods have been developed to evaluate these complex integrals. We introduce the code i-flow, a python package that performs high-dimensional numerical integration utilizing normalizing flows. Normalizing flows are machine-learned, bijective mappings between two distributions. i-flow can also be used to sample random points according to complicated distributions in high dimensions. We compare i-flow to other algorithms for high-dimensional numerical integration and show that i-flow outperforms them for high dimensional correlated integrals. The i-flow code is publicly available on gitlab at https://gitlab.com/i-flow/i-flow.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Machine learning maximized Anderson localization of phonons in aperiodic superlattices

Nanostructuring materials to achieve ultra-low lattice thermal conductivity has proven to be extremely attractive for numerous applications such as thermoelectric energy conversion. Anderson localization of phonons due to aperiodicity can reduce thermal conductivity in superlattices, but the lower limit of thermal conductivity remains elusive due to the prohibitively large design space. In this work, we demonstrate that an intuition-based manual search for aperiodic superlattice structures (random multilayers or RMLs) with the lowest thermal conductivity yields only a local minimum, while a genetic algorithm (GA) based approach can efficiently identify the globally minimum thermal conductivity by only exploring a small fraction of the design space. Our results show that this minimum value occurs at an average RML period that is, surprisingly, smaller than the period corresponding to the minimum SL thermal conductivity. Interestingly, above this critical period, scattering of incoherent phonons at interfaces is less, whereas below this period, the room for randomization becomes less, thus putting more coherent phonons out of Anderson localization and causing increased thermal conductivity. Moreover, the lower limit of the thermal conductivity occurs at a moderate rather than maximum randomness of the layer thickness. Our machine learning approach demonstrates a general process of exploring an otherwise prohibitively large design space to gain non-intuitive physical insights.

42 ENGINEERING↗

SU-D-202-04: Validation of Deformable Image Registration Algorithms for Head and Neck Adaptive Radiotherapy in Routine Clinical Setting

Purpose: To evaluate the ROI contours and accumulated dose difference using different deformable image registration (DIR) algorithms for head and neck (H&N) adaptive radiotherapy. Methods: Eight H&N cancer patients were randomly selected from the affiliated hospital. During the treatment, patients were rescanned every week with ROIs well delineated by radiation oncologist on each weekly CT. New weekly treatment plans were also re-designed with consistent dose prescription on the rescanned CT and executed for one week on Siemens CT-on-rails accelerator. At the end, we got six weekly CT scans from CT1 to CT6 including six weekly treatment plans for each patient. The primary CT1 was set as the reference CT for DIR proceeding with the left five weekly CTs using ANACONDA and MORFEUS algorithms separately in RayStation and the external skin ROI was set to be the controlling ROI both. The entire calculated weekly dose were deformed and accumulated on corresponding reference CT1 according to the deformation vector field (DVFs) generated by the two different DIR algorithms respectively. Thus we got both the ANACONDA-based and MORFEUS-based accumulated total dose on CT1 for each patient. At the same time, we mapped the ROIs on CT1 to generate the corresponding ROIs on CT6 using ANACONDA and MORFEUS DIR algorithms. DICE coefficients between the DIR deformed and radiation oncologist delineated ROIs on CT6 were calculated. Results: For DIR accumulated dose, PTV D95 and Left-Eyeball Dmax show significant differences with 67.13 cGy and 109.29 cGy respectively (Table1). For DIR mapped ROIs, PTV, Spinal cord and Left-Optic nerve show difference with −0.025, −0.127 and −0.124 (Table2). Conclusion: Even two excellent DIR algorithms can give divergent results for ROI deformation and dose accumulation. As more and more TPS get DIR module integrated, there is an urgent need to realize the potential risk using DIR in clinical.

60 APPLIED LIFE SCIENCES↗

The combined force field-sampling problem in simulations of disordered amyloid-β peptides

Molecular dynamics simulations of intrinsically disordered proteins (IDPs) can provide high resolution structural ensembles if the force field is accurate enough and if the simulation sufficiently samples the conformational space of the IDP with the correct weighting of sub-populations. Here, we investigate the combined force field–sampling problem by testing a standard force field as well as newer fixed charge force fields, the latter specifically motivated for better description of unfolded states and IDPs, and comparing them with a standard temperature replica exchange (TREx) protocol and a non-equilibrium Temperature Cool Walking (TCW) sampling algorithm. The force field and sampling combinations are used to characterize the structural ensembles of the amyloid-beta peptides Aβ42 and Aβ43, which both should be random coils as shown recently by experimental nuclear magnetic resonance (NMR) and 2D Förster resonance energy transfer (FRET) experiments. The results illustrate the key importance of the sampling algorithm: while the standard force field using TREx is in poor agreement with the NMR J-coupling and nuclear Overhauser effect and 2D FRET data, when using the TCW method, the standard and optimized protein-water force field combinations are in very good agreement with the same experimental data since the TCW sampling method produces qualitatively different ensembles than TREx. Lastly, we also discuss the relative merit of the 2D FRET data when validating structural ensembles using the different force fields and sampling protocols investigated in this work for small IDPs such as the Aβ42 and Aβ43 peptides.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Estimation of distributions via multilevel Monte Carlo with stratified sampling

We design and implement a novel algorithm for computing a multilevel Monte Carlo (MLMC) estimator of the joint cumulative distribution function (CDF) of a vector-valued quantity of interest in problems with random input parameters and initial conditions. Our approach combines MLMC with stratified sampling of the input sample space by replacing standard Monte Carlo at each level with stratified Monte Carlo initialized with proportionally allocated samples. We show that the resulting stratified MLMC (sMLMC) algorithm is more efficient than its standard MLMC counterpart due to the additional variance reduction provided by the stratification of the random parameter's domain, especially at the coarsest levels. Additional computational cost savings are obtained by smoothing the indicator function with a Gaussian kernel, which proves to be an efficient and robust alternative to recently developed polynomial-based techniques.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Characterizing and mitigating coherent errors in a trapped ion quantum processor using hidden inverses

Quantum computing testbeds exhibit high-fidelity quantum control over small collections of qubits, enabling performance of precise, repeatable operations followed by measurements. Currently, these noisy intermediate-scale devices can support a sufficient number of sequential operations prior to decoherence such that near term algorithms can be performed with proximate accuracy (like chemical accuracy for quantum chemistry problems). While the results of these algorithms are imperfect, these imperfections can help bootstrap quantum computer testbed development. Demonstrations of these algorithms over the past few years, coupled with the idea that imperfect algorithm performance can be caused by several dominant noise sources in the quantum processor, which can be measured and calibrated during algorithm execution or in post-processing, has led to the use of noise mitigation to improve typical computational results. Conversely, benchmark algorithms coupled with noise mitigation can help diagnose the nature of the noise, whether systematic or purely random. Here, we outline the use of coherent noise mitigation techniques as a characterization tool in trapped-ion testbeds. We perform model-fitting of the noisy data to determine the noise source based on realistic physics focused noise models and demonstrate that systematic noise amplification coupled with error mitigation schemes provides useful data for noise model deduction. Further, in order to connect lower level noise model details with application specific performance of near term algorithms, we experimentally construct the loss landscape of a variational algorithm under various injected noise sources coupled with error mitigation techniques. This type of connection enables application-aware hardware codesign, in which the most important noise sources in specific applications, like quantum chemistry, become foci of improvement in subsequent hardware generations.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Using Natural Language Processing of Free-Text Radiology Reports to Identify Type 1 Modic Endplate Changes

Electronic medical record (EMR) systems provide easy access to radiology reports and offer great potential to support quality improvement efforts and clinical research. Harnessing the full potential of the EMR requires scalable approaches such as natural language processing (NLP) to convert text into variables used for evaluation or analysis. Our goal was to determine the feasibility of using NLP to identify patients with Type 1 Modic endplate changes using clinical reports of magnetic resonance (MR) imaging examinations of the spine. Identifying patients with Type 1 Modic change who may be eligible for clinical trials is important as these findings may be important targets for intervention. Four annotators identified all reports that contained Type 1 Modic change, using N = 458 randomly selected lumbar spine MR reports. We then implemented a rule-based NLP algorithm in Java using regular expressions. The prevalence of Type 1 Modic change in the annotated dataset was 10%. Results were recall (sensitivity) 35/50 = 0.70 (95% confidence interval (C.I.) 0.52–0.82), specificity 404/408 = 0.99 (0.97–1.0), precision (positive predictive value) 35/39 = 0.90 (0.75–0.97), negative predictive value 404/419 = 0.96 (0.94–0.98), and F1-score 0.79 (0.43–1.0). Our evaluation shows the efficacy of rule-based NLP approach for identifying patients with Type 1 Modic change if the emphasis is on identifying only relevant cases with low concern regarding false negatives. As expected, our results show that specificity is higher than recall. This is due to the inherent difficulty of eliciting all possible keywords given the enormous variability of lumbar spine reporting, which decreases recall, while availability of good negation algorithms improves specificity.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Computation of Effective Radiative Properties of Powders for Selective Laser Sintering Simulations

Selective laser sintering (SLS) is an additive manufacturing technique for rapidly creating parts directly from a computer-aided design (CAD) model by using a laser to fuse successive layers of powder. However, better understanding of the effect of particle-level variations on the overall build quality is needed. In this work, we investigated these effects computationally by considering the role of the particle size distribution and variations in the powder bed depth to mimic the part complexity found in overhangs and protrusions. In addition, the results from these studies can be distilled to obtain better effective material properties such as laser absorptivity and laser extinction coefficient that are needed for continuum models of the process. We implement a Monte Carlo ray-tracing algorithm within the discrete element model in the open-source simulation software MFiX. Random, loose-packed, particle bed structures are generated, and effective absorptivity and extinction coefficients are calculated. Results are compared against previous computational and experimental measurements for free, monodisperse, and deep powder beds, with good agreement being obtained. Correlations along with uncertainties are developed to allow the effective absorptivity and extinction coefficient as a function of various particle and operational parameters to be accurately set in SLS macroscale models.

Moser, Daniel↗

Compressing unstructured mesh data from simulations using machine learning

The amount of data output from a computer simulation has grown to terabytes and petabytes as increasingly complex simulations are being run on massively parallel systems. As we approach exaflop computing in the next decade, it is expected that the I/O subsystem will not be able to write out these large volumes of data. In this paper, we explore the use of machine learning to compress the data before it is written out. Despite the computational constraints that limit us to using very simple learning algorithms, our results show that machine learning is a viable option for compressing unstructured data. Furthermore, we demonstrate that by simply using a better sampling algorithm to generate the training set, we can obtain more accurate results compared to random sampling, but at no extra cost. Further, by carefully selecting and incorporating points with high prediction error, we can improve reconstruction accuracy without sacrificing the compression rate.

97 MATHEMATICS AND COMPUTING↗

A mapped dataset of surface ocean acidification indicators in large marine ecosystems of the United States

Mapped monthly data products of surface ocean acidification indicators from 1998 to 2022 on a 0.25° by 0.25° spatial grid have been developed for eleven U.S. large marine ecosystems (LMEs). The data products were constructed using observations from the Surface Ocean CO 2 Atlas, co-located surface ocean properties, and two types of machine learning algorithms: Gaussian mixture models to organize LMEs into clusters of similar environmental variability and random forest regressions (RFRs) that were trained and applied within each cluster to spatiotemporally interpolate the observational data. The data products, called RFR-LMEs, have been averaged into regional timeseries to summarize the status of ocean acidification in U.S. coastal waters, showing a domain-wide carbon dioxide partial pressure increase of 1.4 ± 0.4 μatm yr -1 and pH decrease of 0.0014 ± 0.0004 yr -1 . RFR-LMEs have been evaluated via comparisons to discrete shipboard data, fixed timeseries, and other mapped surface ocean carbon chemistry data products. Regionally averaged timeseries of RFR-LME indicators are provided online through the NOAA National Marine Ecosystem Status web portal.

54 ENVIRONMENTAL SCIENCES↗

How Generalizable is a Machine-Learning Approach for Modeling Hub-Height Turbulence Intensity?

Hub-height turbulence intensity is essential for a variety of wind energy applications. However, simulating it is a challenging task. Simple analytical models have been proposed in the literature, but they all come with significant limitations. Even state-of-the-art numerical weather prediction models, such as the Weather Research and Forecasting model, currently struggle to predict hub-height turbulence intensity. Here, we propose a machine-learning-based approach to predict hub-height turbulence intensity from other hub-height and ground-level atmospheric measurements, using observations from the Perdigao field campaign and the Southern Great Plains atmospheric observatory. We consider a random forest regression model, which we validate first at the site used for training and then under a more robust round-robin approach, and compare its performance to a multivariate linear regression. The random forest successfully outperforms the linear regression in modeling hub-height turbulence intensity, with a normalized root-mean-square error as low as 0.014 when using 30-minute average data. In order to achieve such low root-mean-square error values, the knowledge of hub-height turbulence kinetic energy (which can instead be modeled in the Weather Research and Forecasting model) is needed. Interestingly, we find that the performance of the random forest generalizes well when considering a round-robin validation (i.e., when the algorithm is trained at one site such as Perdigao or Southern Great Plains) and then applied to model hub-height turbulence intensity at the other location.

17 WIND ENERGY↗

Metropolis-style random sampling of quantum gates for the estimation of low-energy observables

In this work, we propose a quantum algorithm to compute low-energy expectation values of a quantum Hamiltonian by sampling a partition function associated with the average energy of that Hamiltonian. For any given quantum circuit-Hamiltonian pair, there is an associated average energy. The sampling is done through an accept/reject Metropolis-style algorithm on the quantum gates of the circuit itself. Observables calculated under the canonical ensemble from these samples of circuits are extrapolated from higher energies to the ground state.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Evolving Larger Convolutional Layer Kernel Sizes for a Settlement Detection Deep-Learner on Summit

Deep-learner hyper-parameters, such as kernel sizes, batch sizes, and learning rates, can significantly influence the quality of trained models. The state of the art for finding optimal hyper-parameters generally uses a brute force, grid search approach, random search, or Bayesian-based optimization among other techniques. We applied an evolutionary algorithm to optimize kernel sizes for a convolutional neural network used to detect settlements in satellite imagery. Usually convolutional layer kernel sizes are small - typically one, three, or five - but we found that the system converged at, or near, kernel sizes of nine for the last convolutional layer, and that this occurred for multiple runs using two different datasets. Moreover, the larger kernel sizes had fewer false positives than the 3x3 kernel sizes found as optimal via a brute force uniform grid search. This suggests that this large kernel size may be leveraging patterns found in larger areal features in the source imagery, and that this may be generalized as possible guidance for similar remote sensing deep-learning tasks.

Coletti, Mark↗

Monte Carlo investigation of the increased radiation deposition due to gold nanoparticles using kilovoltage and megavoltage photons in a 3D randomized cell model

Purpose: Investigation of increased radiation dose deposition due to gold nanoparticles (GNPs) using a 3D computational cell model during x-ray radiotherapy.Methods: Two GNP simulation scenarios were set up in Geant4; a single 400 nm diameter gold cluster randomly positioned in the cytoplasm and a 300 nm gold layer around the nucleus of the cell. Using an 80 kVp photon beam, the effect of GNP on the dose deposition in five modeled regions of the cell including cytoplasm, membrane, and nucleus was simulated. Two Geant4 physics lists were tested: the default Livermore and custom built Livermore/DNA hybrid physics list. 10{sup 6} particles were simulated at 840 cells in the simulation. Each cell was randomly placed with random orientation and a diameter varying between 9 and 13 {mu}m. A mathematical algorithm was used to ensure that none of the 840 cells overlapped. The energy dependence of the GNP physical dose enhancement effect was calculated by simulating the dose deposition in the cells with two energy spectra of 80 kVp and 6 MV. The contribution from Auger electrons was investigated by comparing the two GNP simulation scenarios while activating and deactivating atomic de-excitation processes in Geant4.Results: The physical dose enhancement ratio (DER) of GNP was calculated using the Monte Carlo model. The model has demonstrated that the DER depends on the amount of gold and the position of the gold cluster within the cell. Individual cell regions experienced statistically significant (p < 0.05) change in absorbed dose (DER between 1 and 10) depending on the type of gold geometry used. The DER resulting from gold clusters attached to the cell nucleus had the more significant effect of the two cases (DER {approx} 55). The DER value calculated at 6 MV was shown to be at least an order of magnitude smaller than the DER values calculated for the 80 kVp spectrum. Based on simulations, when 80 kVp photons are used, Auger electrons have a statistically insignificant (p < 0.05) effect on the overall dose increase in the cell. The low energy of the Auger electrons produced prevents them from propagating more than 250-500 nm from the gold cluster and, therefore, has a negligible effect on the overall dose increase due to GNP.Conclusions: The results presented in the current work show that the primary dose enhancement is due to the production of additional photoelectrons.

61 RADIATION PROTECTION AND DOSIMETRY↗