Search NASA⌕ Search

SEARCH · Search NASA

Results for “Ensemble computing.”

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 1,045 records · Page 58

EFFECT OF INCLINED WAVES ON DEEPLY EMBEDDED NUCLEAR FACILITIES

Several conceptual designs for advanced reactors show that most if not all of the structure is embedded in the soil. Consequently, analysis techniques that have been verified and validated for the analysis of light water reactors on surface foundations may not be directly applicable to deeply embedded reactors. Specifically two effects that may change the in-structure response (ISR) of embedded facilities are: (i) the effect of the kinematic interaction becomes more prominent as the embedment depth increases; (ii) non-vertically propagating waves, which are usually not considered for surface-founded facilities, may be more important for deeply embedded facilities. A preliminary study is presented in this article to understand the response of embedded nuclear facilities to non-vertically propagating seismic waves. The inclined wave field is generated from source-to-site simulations using representative earthquake faults with different dip angles. The embedment depth of the nuclear containment facility is also varied to obtain a better understanding of the effect of embedment. This analysis of the soil-structure ensemble is conducted using the application called Multi-hazard Analysis for STOchastic time-DOmaiN phenomena (MASTODON), developed on top of Idaho National Lab’s open source MOOSE finite element framework. The Domain Reduction Method implemented in MASTODON to incorporate the seismic excitation within the region of interest is verified and then used to reduce the computational effort in analyzing this problem. Preliminary analysis using linear soil and structure material properties and tied contact between the soil and structure indicate that embedding the nuclear facility reduces significantly the response of the structure in all the earthquake scenarios considered.

42 ENGINEERING↗

A parametrically programmable delay line for microwave photons

Delay lines that store quantum information are crucial for advancing quantum repeaters and hardware efficient quantum computers. Traditionally, they are realized as extended systems that support wave propagation but provide limited control over the propagating fields. Here, we introduce a parametrically addressed delay line for microwave photons that provides a high level of control over the stored pulses. By parametrically driving a three-wave mixing circuit element that is weakly hybridized with an ensemble of resonators, we engineer a spectral response that simulates that of a physical delay line, while providing fast control over the delay line’s properties. We demonstrate this novel degree of control by choosing which photon echo to emit, translating pulses in time, and even swapping two pulses, all with pulse energies on the order of a single photon. We also measure the noise added from our parametric interactions and find it is much less than one photon.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Single-pixel interior filling function approach for detecting and correcting errors in particle tracking

We present a general method for detecting and correcting biases in the outputs of particle-tracking experiments. Our approach is based on the histogram of estimated positions within pixels, which we term the single-pixel interior filling function (SPIFF). We use the deviation of the SPIFF from a uniform distribution to test the veracity of tracking analyses from different algorithms. Unbiased SPIFFs correspond to uniform pixel filling, whereas biased ones exhibit pixel locking, in which the estimated particle positions concentrate toward the centers of pixels. Although pixel locking is a well-known phenomenon, we go beyond existing methods to show how the SPIFF can be used to correct errors. Here, the key is that the SPIFF aggregates statistical information from many single-particle images and localizations that are gathered over time or across an ensemble, and this information augments the single-particle data. We explicitly consider two cases that give rise to significant errors in estimated particle locations: undersampling the point spread function due to small emitter size and intensity overlap of proximal objects. In these situations, we show how errors in positions can be corrected essentially completely with little added computational cost. Additional situations and applications to experimental data are explored in SI Appendix. In the presence of experimental-like shot noise, the precision of the SPIFF-based correction achieves (and can even exceed) the unbiased Cramér–Rao lower bound. We expect the SPIFF approach to be useful in a wide range of localization applications, including single-molecule imaging and particle tracking, in fields ranging from biology to materials science to astronomy.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Transversity GPDs of the proton from lattice QCD

We present the first calculation of the x dependence of the isovector transversity generalized parton distributions (GPDs) for the proton within lattice QCD. We compute the matrix elements with nonlocal operators containing a Wilson line. The calculation implements the Breit symmetric frame. The proton momenta are chosen as 0.83, 1.25, and 1.67 GeV, and the values of the momentum transfer squared are 0.69 and 1.02 GeV 2 . These combinations include cases with zero and nonzero skewness. The calculation is performed using one ensemble of twisted-mass fermions with a clover term: two degenerate-mass light quarks, a strange quark, and a charm quark. The lattice results are renormalized nonperturbatively and finally matched to the light-cone GPDs using one-loop perturbation theory within the framework of large-momentum effective theory. The final GPDs are given in the $\overline{MS}$ scheme at a scale of 2 GeV. In addition to the individual GPDs, we form the combination of the transversity GPDs that is related to the transverse spin structure of the proton. Finally, we extract the lowest two moments of the GPDs and draw a number of important qualitative conclusions.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Uncertainty quantification for neural network potential foundation models

Abstract For neural network potentials (NNPs) to gain widespread use, researchers must be able to trust model outputs. However, the blackbox nature of neural networks and their inherent stochasticity are often deterrents, especially for foundation models trained over broad swaths of chemical space. Uncertainty information provided at the time of prediction can help reduce aversion to NNPs. In this work, we detail two uncertainty quantification (UQ) methods. Readout ensembling, by finetuning the readout layers of an ensemble of foundation models, provides information about model uncertainty, while quantile regression, by replacing point predictions with distributional predictions, provides information about uncertainty within the underlying training data. We demonstrate our approach with the MACE-MP-0 model, applying UQ to the foundation model and a series of finetuned models. The uncertainties produced by the readout ensemble and quantile methods are demonstrated to be distinct measures by which the quality of the NNP output can be judged.

36 MATERIALS SCIENCE↗

Semi-supervised Bayesian Low-shot Learning

Deep neural networks (NNs) typically outperform traditional machine learning (ML) approaches for complicated, non-linear tasks. It is expected that deep learning (DL) should offer superior performance for the important non-proliferation task of predicting explosive device configuration based upon observed optical signature, a task which human experts struggle with. However, supervised machine learning is difficult to apply in this mission space because most recorded signatures are not associated with the corresponding device description, or “truth labels.” This is challenging for NNs, which traditionally require many samples for strong performance. Semi-supervised learning (SSL), low-shot learning (LSL), and uncertainty quantification (UQ) for NNs are emerging approaches that could bridge the mission gaps of few labels and rare samples of importance. NN explainability techniques are important in gaining insight into the inferential feature importance of such a complex model. In this work, SSL, LSL, and UQ are merged into a single framework, a significant technical hurdle not previously demonstrated. Exponential Average Adversarial Training (EAAT) and Pairwise Neural Networks (PNNs) are chosen as the SSL and LSL methods of choice. Permutation feature importance (PFI) for functional data is used to provide explainability via the Variable importance Explainable Elastic Shape Analysis (VEESA) pipeline. A variety of uncertainty quantification approaches are explored: Bayesian Neural Networks (BNNs), ensemble methods, concrete dropout, and evidential deep learning. Two final approaches, one utilizing ensemble methods and one utilizing evidential learning, are constructed and compared using a well-quantified synthetic 2D dataset along with the DIRSIG Megascene.

97 MATHEMATICS AND COMPUTING↗

Two-pion contribution to the hadronic vacuum polarization with staggered quarks

We present results from the first lattice QCD calculation of the two-pion contributions to the light-quark connected vector-current correlation function obtained from staggered-quark operators. We employ the MILC Collaboration’s gauge-field ensemble with 2 + 1 + 1 flavors of highly improved staggered sea quarks at a lattice spacing of a ≈ 0.15 fm with a light sea-quark mass at its physical value. The two-pion contributions allow for a refined determination of the noisy long-distance tail of the vector-current correlation function, which we use to compute the light-quark connected contribution to hadronic vacuum polarization (HVP) with improved statistical precision. We compare our results with traditional noise-reduction techniques used in lattice QCD calculations of the light-quark connected HVP, namely, the so-called fit and bounding methods. We observe a factor of roughly 3 improvement in the statistical precision in the determination of the HVP contribution to the muon’s anomalous magnetic moment over these approaches. We also lay the group theoretical groundwork for extending this calculation to finer lattice spacings with increased numbers of staggered two-pion taste states.

Lahert, Shaun [Utah U.; Illinois U., Urbana] (ORCI↗

Scalable learning of potentials to predict time-dependent Hartree–Fock dynamics

We propose a framework to learn the time-dependent Hartree–Fock (TDHF) inter-electronic potential of a molecule from its electron density dynamics. Although the entire TDHF Hamiltonian, including the inter-electronic potential, can be computed from first principles, we use this problem as a testbed to develop strategies that can be applied to learn a priori unknown terms that arise in other methods/approaches to quantum dynamics, e.g., emerging problems such as learning exchange–correlation potentials for time-dependent density functional theory. We develop, train, and test three models of the TDHF inter-electronic potential, each parameterized by a four-index tensor of size up to 60 × 60 × 60 × 60. Two of the models preserve Hermitian symmetry, while one model preserves an eight-fold permutation symmetry that implies Hermitian symmetry. Across seven different molecular systems, we find that accounting for the deeper eight-fold symmetry leads to the best-performing model across three metrics: training efficiency, test set predictive power, and direct comparison of true and learned inter-electronic potentials. All three models, when trained on ensembles of field-free trajectories, generate accurate electron dynamics predictions even in a field-on regime that lies outside the training set. To enable our models to scale to large molecular systems, we derive expressions for Jacobian-vector products that enable iterative, matrix-free training.

97 MATHEMATICS AND COMPUTING↗

Direction-dependent localization errors in SPECT images

Purpose: Single photon emission computed tomography (SPECT) is being investigated for imaging inside radiation therapy treatment rooms to localize biological targets. Here, computer simulations were used to analyze locational and directional dependencies in localization errors and to assess the effects of spatial resolution modeling and observer normalization on localization performance. Methods: SPECT images of the XCAT phantom, containing 12 hot tumors, were reconstructed with detector response function compensation (DRC) and without DRC (nDRC). Numerical observers were forced to select the most suspicious tumor location, using normalized cross correlation (NXC) or un-normalized cross correlation (XC), from 3 cm diameter search volumes that each contained only one tumor. For each tumor site, localization was optimized as a function of the iteration number and postreconstruction smoothing. Localization error, the distance between true and estimated tumor positions, was calculated across the ensembles of 80 images. Direction-dependent localization bias and precision were estimated from the image ensemble. Results: For the six superficial tumors in close proximity to the detector trajectory, mean localization errors were <2 mm and were lowest or comparable using DRC-NXC, though differences from DRC-XC and nDRC-NXC were not statistically significant. DRC-NXC did provide statistically significantly better localization than nDRC-XC for five of these six tumors. At the other six sites where attenuation was more severe and the distance was generally greater between the tumor and detector, DRC typically did not show better localization than nDRC. Observer normalization improved the localization substantially for a tumor near the hotter heart. Localization errors were anisotropic and dependent on tumor location relative to the detector trajectory. Conclusions: This computer-simulation study compared localization performance for normalized and un-normalized numerical observers, which were used to estimate tumor positions in SPECT images, reconstructed with and without DRC. For tumors localized to <2 mm on average, which are good candidates for SPECT-guided radiation therapy, localization performance typically improved by compensating for the detector response function and by using a normalized observer. The observed direction-dependent localization errors have important implications for radiation therapy and are relevant to SPECT imaging in general.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Lattice QCD at non-zero isospin chemical potential

Systems of non-zero isospin chemical potential are studied from a canonical approach by computing correlation functions with the quantum numbers of N π(+)&#x27;s (C(N)(π)). In order to reduce the number of contractions required in calculating C(N)(π) for a large N in the Wick&#x27;s theorem, we constructed a few new algorithms. With these new algorithms, systems with isospin charge up to 72 are investigated on three anisotropic gauge ensembles with a pion mass of 390 MeV, and with lattice spatial extents L ~ 2.0, 2.5, 3.0 fm. The largest isospin density of ρ(I) thickapprox 9 fm(-)(3) is achieved in the smallest volume, and the QCD phase diagram is investigated at a fixed low temperature at varying isospin chemical potentials, m(π) ≤ μ(I) ≤ 4.5 m(π). By investigating the behaviour of the extracted energy density of the system at different isospin chemical potentials, we numerically identified the conjectured transition to a Bose-Einstein condensation state at μ(I) ≥ m(π).

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

SPICE: discovery of phenotype-determining component interplays

Background: A latent behavior of a biological cell is complex. Deriving the underlying simplicity, or the fundamental rules governing this behavior has been the Holy Grail of systems biology. Data-driven prediction of the system components and their component interplays that are responsible for the target system’s phenotype is a key and challenging step in this endeavor. Results: The proposed approach, which we call System Phenotype-related Interplaying Components Enumerator (SPICE), iteratively enumerates statistically significant system components that are hypothesized (1) to play an important role in defining the specificity of the target system’s phenotype(s); (2) to exhibit a functionally coherent behavior, namely, act in a coordinated manner to perform the phenotype-specific function; and (3) to improve the predictive skill of the system’s phenotype(s) when used collectively in the ensemble of predictive models. SPICE can be applied to both instance-based data and network-based data. When validated, SPICE effectively identified system components related to three target phenotypes: biohydrogen production, motility, and cancer. Manual results curation agreed with the known phenotype-related system components reported in literature. Additionally, using the identified system components as discriminatory features improved the prediction accuracy by 10% on the phenotype-classification task when compared to a number of state-of-the-art methods applied to eight benchmark microarray data sets. Conclusion: We formulate a problem—enumeration of phenotype-determining system component interplays—and propose an effective methodology (SPICE) to address this problem. SPICE improved identification of cancer-related groups of genes from various microarray data sets and detected groups of genes associated with microbial biohydrogen production and motility, many of which were reported in literature. SPICE also improved the predictive skill of the system’s phenotype determination compared to individual classifiers and/or other ensemble methods, such as bagging, boosting, random forest, nearest shrunken centroid, and random forest variable selection method.

59 BASIC BIOLOGICAL SCIENCES↗

Understanding and control of Zener pinning via phase field and ensemble learning

Zener pinning refers to the dispersion of fine particles which influences grain size distribution via movement of grain boundaries in a polycrystalline material. Grain size distribution in polycrystals has a significant impact on their properties including physical, chemical, mechanical, and optical to name a few. We explore the use of Phase-field modeling and machine-learning techniques to understand and improve the control of grain size distribution via Zener pinning in polycrystalline materials. We develop a machine learning model that determines the relative importance of various parameters to exercise microstructure control via Zener pinning. Our workflow combines high-throughput phase-field simulations and machine learning to address the computational bottlenecks associated with large-scale simulations as well as identify features necessary for microstructure control in polycrystals. A random forest (RF) regression model was developed to predict grain sizes based on five Phase-field model parameters, achieving an average prediction error of 0.72 nm for the training data and 1.44 nm for the test data. The importance of the input parameters is analyzed using the SHapley Additive exPlanations (SHAP) approach which reveals that diffusivity, volume fraction, and particle diameter are the most important parameters in determining the final grain size. These findings will allow us to select the best second-phase particles, optimize grain size distributions and thus design microstructures with the desired properties. The developed method is a highly versatile and generalizable approach that can be used to assess the combined effects of individual features in the presence of multiple variables.

36 MATERIALS SCIENCE↗

Analyzing and Exploring Training Recipes for Large-Scale Transformer-Based Weather Prediction

Abstract The rapid rise of deep learning (DL) in numerical weather prediction (NWP) has led to a proliferation of models which forecast atmospheric variables with comparable or superior skill than traditional physics-based NWP. However, among these leading DL models, there is a wide variance in both the training settings and architecture used. Further, the lack of thorough ablation studies makes it hard to discern which components are most critical to success. In this work, we show that it is possible to attain high forecast skill even with relatively off-the-shelf architectures, simple training procedures, and moderate compute budgets. Specifically, we train a minimally modified Swin Transformer V2 (SwinV2) on ERA5 data and find that it attains superior skill in terms of mean-square errors of deterministic forecasts when compared against the European Centre for Medium-Range Weather Forecasts’ Integrated Forecasting System (IFS). Almost all DL–NWP systems share a core set of hyperparameters and design decisions. To aid and expedite future DL–NWP research, we present an in-depth, systematic exploration of different loss functions, model sizes and depths, patch sizes, and multistep training objectives. We also examine the model performance with metrics beyond the typical accuracy (ACC) and RMSE and investigate how the performance scales with model size. Through our open-source code, scoring pipelines, and models, we share our findings on key aspects of the training pipeline. These ablations reduce the necessity for expensive hyperparameter tuning and lower the barrier to entry for future DL–NWP research. Significance Statement This study investigates the potential of using large-scale transformer-based models for weather prediction, showing that it is possible to achieve high forecast accuracy with simpler, off-the-shelf architectures. By training a minimally modified SwinV2 transformer on ERA5 data, we show that the model achieves competitive forecast skill in terms of mean-square error for key variables, outperforming the European Centre for Medium-Range Weather Forecasts’ Integrated Forecasting System (IFS) at all lead times. Our findings suggest that effective training strategies, such as multistep fine-tuning and channel-weighted losses, significantly enhance the model’s performance. However, we also highlight that these improvements come with trade-offs in other areas, such as ensemble spread and high-frequency spatial detail. This work highlights the promise of deep learning in improving weather forecasts, which could lead to better preparedness and response to weather events, ultimately benefiting society by providing more reliable weather predictions.

Willard, Jared D. [Lawrence Berkeley National Labo↗

Robust finite-temperature many-body scarring on a quantum computer

Mechanisms for suppressing thermalization in disorder-free many-body systems, such as Hilbert space fragmentation and quantum many-body scars, have recently attracted much interest in foundations of quantum statistical physics and potential quantum information processing applications. However, their sensitivity to realistic effects such as finite temperature remains largely unexplored. Here, we have utilized IBM's Kolkata quantum processor to demonstrate an unexpected robustness of quantum many-body scars at finite temperatures when the system is prepared in a thermal Gibbs ensemble. We identify such robustness in the PXP model, which describes quantum many-body scars in experimental systems of Rydberg atom arrays and ultracold atoms in tilted Bose-Hubbard optical lattices. By contrast, other theoretical models which host exact quantum many-body scars are found to lack such robustness and their scarring properties quickly decay with temperature. Our study sheds light on the important differences between scarred models in terms of their algebraic structures, which impacts their resilience to finite temperature. Published by the American Physical Society 2024

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Replica wormholes and the black hole interior

Recent work has shown how to obtain the Page curve of an evaporating black hole from holographic computations of entanglement entropy. We show how these computations can be justified using the replica trick, from geometries with a spacetime wormhole connecting the different replicas. In a simple model, we study the Page transition in detail by summing replica geometries with different topologies. We compute related quantities in less detail in more complicated models, including JT gravity coupled to conformal matter and the SYK model. Separately, we give a direct gravitational argument for entanglement wedge reconstruction using an explicit formula known as the Petz map; again, a spacetime wormhole plays an important role. We discuss an interpretation of the wormhole geometries as part of some ensemble average implicit in the gravity description.

2D gravity↗