Quantum Privacy Aggregation of Teacher Ensembles (QPATE) for Privacy Preserving Quantum Machine Learning
Explore the source record for details and available documents.
SEARCH · Search NASA
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.
Explore the source record for details and available documents.
Behavior of threading dislocations in porous heteroepitaxial gallium nitride (GaN) films has been studied using computer simulation by the two-dimensional discrete dislocation dynamics approach. A computational scheme, where pores are modeled as cross sections of cylindrical cavities, elastically interacting with unidirectional parallel edge dislocations, which imitate threading dislocations, is used. Time dependences of coordinates and velocities of each dislocation from dislocation ensembles under investigation are obtained. Visualization of current structure of dislocation ensemble is performed in the form of a location map of dislocations at any time. It has been shown that the density of appearing dislocation structures significantly depends on the ratio of area of a pore cross section to area of the simulation region. In particular, increasing the portion of pores surface on the layer surface up to 2% should lead to about a 1.5-times decrease of the final density of threading dislocations, and increase of this portion up to 15% should lead to approximately a 4.5-times decrease of it.
A fiber array polar nephelometer of advanced design, the FAN I is capable of in-situ phase function measurements of scattered light from man-made or natural atmospheric particles. The scattered light is measured at 100 different angles throughout 360 degrees, thus providing a potential measurement of the asymmetry of irregularly shaped particles. Phase functions can be measured at 10 to 100 Hz rates and the range of measurable single particle sizes is from 5 micron m to as large as 8mm. For particles smaller than 5 micro m the ensemble average can be measured. The FAN I is microprocessor controlled and the data may be stored on floppy disk or printed out in tabular and/or graphical form. The optical head may be separated from the computer system for operation in field or adverse conditions. Examples of laboratory measured scattering phase functions obtained with the FAN I for spherical particles is given to illustrate its measurement capabilities.
A tight binding model which includes s, p, d, s orbitals is used to examine the electronic structures of an ensemble of dome-shaped In0.6 Ga0.4 As quantum dots. Given ensembles of identically sized quantum dots, variations in composition and configuration yield a linewidth broadening of less than 0.35 meV, much smaller than the total broadening determined from photoluminescence experiments. It is also found that the computed disorder-induced broadening is very sensitive to the applied boundary conditions, so that care must be taken to ensure proper convergence of the numerical results. Examination of local eigenenergies as functions of position shows similar convergence problems and indicates that an inaccurate resolution of the equilibrium atomic positions due to truncation of the simulation domain may be the source of the slow ground state convergence.
Subglacial drainage models represent water flow at the ice–bed interface through coupled distributed and channelized systems to determine water pressure, discharge, and drainage system geometry. While they are used to understand processes such as the relationship between surface melt and ice flow, the number of uncertain model parameters and the computational cost of running models makes it difficult to adequately explore the high-dimensional parameter space and evaluate uncertainty in model predictions. Here, we develop Gaussian process (GP) emulators that make fast predictions with associated uncertainty of subglacial drainage model outputs. Using a truncated principal component (PC) basis representation, we construct a GP emulator for diurnally averaged subglacial water pressure. We also explore emulation of scalar variables describing drainage efficiency and configuration. We train the emulators using ensembles of up to 512 simulations varying eight parameters of the Glacier Drainage System (GlaDS) model on a synthetic domain intended to represent an ice-sheet margin. The emulators make predictions ∼ 1000 times faster than GlaDS simulations, with errors <3 % for the water pressure field and ∼ 5 %–9 % for drainage efficiency and configuration. We apply the emulators to explore the eight-dimensional parameter space by computing variance-based parameter sensitivity indices, finding that three parameters (ice flow coefficient, bed bump aspect ratio, and the subglacial cavity system conductivity) explain 90 % of the variance in modelled water pressure in response to parameter changes. The GP emulator approach described here is well suited to integrating observational data with models to make calibrated, credible predictions of subglacial drainage.
Not provided.
Background: As healthcare increasingly digitizes, streaming waveform data is being made available from an variety of sources, but there still remains a paucity of performant clinical decision support systems. For example, in the intensive care unit (ICU) existing automated alarm systems typically rely on simple thresholding that result in frequent false positives. Recurrent false positive alerts create distrust of alarm mechanisms that can be directly detrimental to patient health. To improve patient care in the ICU, we need alert systems that are both pervasive, and accurate so as to be informative and trusted by providers. Objective: We aimed to develop a machine learning-based classifier to detect abnormal waveform events using the use case of mechanical ventilation waveform analysis, and the detection of harmful forms of ventilation delivery to patients. We specifically focused on detecting injurious subtypes of patient-ventilator asynchrony (PVA).Methods: Using a dataset of breaths recorded from 35 different patients, we used machine learning to create computational models to automatically detect, and classify two types of injurious PVA, double trigger asynchrony (DTA), breath stacking asynchrony (BSA). We examined the use of synthetic minority over-sampling technique (SMOTE) to overcome class imbalance problems, varied methods for feature selection, and use of ensemble methods to optimize the performance of our model. Results: We created an ensemble classifier that is able to accurately detect DTA at a sensitivity/specificity of 0.960/0.975, BSA at sensitivity/specificity of 0.944/0.987, and non-PVA events at sensitivity/specificity of .967/.980. Conclusions: Our results suggest that it is possible to create a high-performing machine learning-based model for detecting PVA in mechanical ventilator waveform data in spite of both intra-patient, and inter-patient variability in waveform patterns, and the presence of clinical artifacts like cough and suction procedures. Our work highlights the importance of addressing class imbalance in clinical data sets, and the combined use of statistical methods and expert knowledge in feature selection.
Using equilibrium fluctuations to understand the response of a physical system to an externally imposed perturbation is the basis for linear response theory, which is widely used to interpret experiments and shed light on microscopic dynamics. For nonequilibrium systems, perturbations cannot be interpreted simply by monitoring fluctuations in a conjugate observable and general response results rely on path ensemble averaging. Furthermore, these techniques do not apply to perturbations that affect the diffusion tensor in a stochastic system. Here, we introduce an “effective” physical process that represents the diffusion perturbed dynamics and enables accurate calculations of responses to a change in the diffusion. Interestingly, the effective dynamics contain an additional drift involving the instantaneous “score” of the system, and we leverage score matching algorithms to carry out nonequilibrium response calculations on systems for which the exact stationary distribution is unknown.
We present a model-assisted density fitting (MADF) basis set generator, an algorithm for generating primitive atomic Gaussian density fitting (DF) basis sets (DFBSs) from a contracted Gaussian orbital basis set (OBS). The MADF algorithm produces DFBSs suitable for accurate robust DF approximation of 2-particle interactions in mean-field and correlated electronic structures. The algorithm is designed to (a) saturate the OBS product space by a large regularized set of primitive solid-harmonic Gaussian shells with nonuniform distribution of exponents, followed by (b) pruning of the shells according to their contributions to the 2- body energy of a correlated atomic ensemble. Building the DFBS generator model almost exclusively on mathematical and physical principles allows one to limit the number of parameters that control the density fitting error to three, with a single set of parameters sufficient for computations with all basis cardinal numbers, with and without correlation of core electrons, with and without scalar and spin-dependent relativistic effects, spanning almost all of the Periodic Table. Performance assessment included basis sets up to quadruple-ζ quality from several major basis set families, using molecules composed of main-group, d-block, and f-block elements. The resulting DF errors in Hartree−Fock and second-order MP2 energies (with relativistic all-electron treatments, when appropriate) were on the order of 20 and 10 μE h per electron, respectively.
We have performed large-eddy simulations of decaying isotropic turbulence to test the prediction of self-similar decay of the energy spectrum and to compute the decay exponents of the kinetic energy. In general, good agreement between the simulation results and the assumption of self-similarity were obtained. However, the statistics of the simulations were insufficient to compute the value of gamma which corrects the decay exponent when the spectrum follows a k(exp 4) wave number behavior near k = 0. To obtain good statistics, it was found necessary to average over a large ensemble of turbulent flows.
High-frequency vibroacoustic modeling is typically performed using energy-based techniques such as Statistical Energy Analysis (SEA). Energy models require an estimate of the internal damping loss factor. Unfortunately, the loss factor is difficult to estimate analytically, and experimental methods such as the power injection method can require extensive measurements over the structure of interest. This paper discusses the implications of estimating damping loss factors using the impulse response decay method (IRDM) from a limited set of response measurements. An automated procedure for implementing IRDM is described and then evaluated using data from a finite element model of a stiffened, curved panel. Estimated loss factors are compared with loss factors computed using a power injection method and a manual curve fit. The paper discusses the sensitivity of the IRDM loss factor estimates to damping of connected subsystems and the number and location of points in the measurement ensemble.
Hyperspectral data are becoming more widely available via sensors on airborne and unmanned aerial vehicle (UAV) platforms, as well as proximal platforms. While space-based hyperspectral data continue to be limited in availability, multiple spaceborne Earth-observing missions on traditional platforms are scheduled for launch, and companies are experimenting with small satellites for constellations to observe the Earth, as well as for planetary missions. Land cover mapping via classification is one of the most important applications of hyperspectral remote sensing and will increase in significance as time series of imagery are more readily available. However, while the narrow bands of hyperspectral data provide new opportunities for chemistry-based modeling and mapping, challenges remain. Hyperspectral data are high dimensional, and many bands are highly correlated or irrelevant for a given classification problem. For supervised classification methods, the quantity of training data is typically limited relative to the dimension of the input space. The resulting Hughes phenomenon, often referred to as the curse of dimensionality, increases potential for unstable parameter estimates, overfitting, and poor generalization of classifiers. This is particularly problematic for parametric approaches such as Gaussian maximum likelihood–based classifiers that have been the backbone of pixel-based multispectral classification methods. This issue has motivated investigation of alternatives, including regularization of the class covariance matrices, ensembles of weak classifiers, development of feature selection and extraction methods, adoption of nonparametric classifiers, and exploration of methods to exploit unlabeled samples via semi-supervised and active learning. Data sets are also quite large, motivating computationally efficient algorithms and implementations. This chapter provides an overview of the recent advances in classification methods for mapping vegetation using hyperspectral data. Three data sets that are used in the hyperspectral classification literature (e.g., Botswana Hyperion satellite data and AVIRIS airborne data over both Kennedy Space Center and Indian Pines) are described in Section 3.2 and used to illustrate methods described in the chapter. An additional high-resolution hyperspectral data set acquired by a SpecTIR sensor on an airborne platform over the Indian Pines area is included to exemplify the use of new deep learning approaches, and a multiplatform example of airborne hyperspectral data is provided to demonstrate transfer learning in hyperspectral image classification. Classical approaches for supervised and unsupervised feature selection and extraction are reviewed in Section 3.3. In particular, nonlinearities exhibited in hyperspectral imagery have motivated development of nonlinear feature extraction methods in manifold learning, which are outlined in Section 3.3.1.4. Spatial context is also important in classification of both natural vegetation with complex textural patterns and large agricultural fields with significant local variability within fields. Approaches to exploit spatial features at both the pixel level (e.g., co-occurrence–based texture and extended morphological attribute profiles [EMAPs]) and integration of segmentation approaches (e.g., HSeg) are discussed in this context in Section 3.3.2. Recently, classification methods that leverage nonparametric methods originating in the machine learning community have grown in popularity. An overview of both widely used and newly emerging approaches, including support vector machines (SVMs), Gaussian mixture models, and deep learning based on convolutional neural networks is provided in Section 3.4. Strategies to exploit unlabeled samples, including active learning and metric learning, which combine feature extraction and augmentation of the pool of training samples in an active learning framework, are outlined in Section 3.5. Integration of image segmentation with classification to accommodate spatial coherence typically observed in vegetation is also explored, including as an integrated active learning system. Exploitation of multisensor strategies for augmenting the pool of training samples is investigated via a transfer learning framework in Section 3.5.1.2. Finally, we look to the future, considering opportunities soon to be provided by new paradigms, as hyperspectral sensing is becoming common at multiple scales from ground-based and airborne autonomous vehicles to manned aircraft and space-based platforms.
The systems and methods described herein utilize a probabilistic modeling framework for reverse engineering an ensemble of causal models, from data and then forward simulating the ensemble of models to analyze and predict the behavior of the network. In certain embodiments, the systems and methods described herein include data-driven techniques for developing causal models for biological networks. Causal network models include computational representations of the causal relationships between independent variables such as a compound of interest and dependent variables such as measured DNA alterations, changes in mRNA, protein, and metabolites to phenotypic readouts of efficacy and toxicity.
We compute the leptonic decay constants $f_{D^+}$, $f_{D_s}$, and $f_{K^+}$, and the quark-mass ratios $m_c/m_s$ and $m_s/m_l$ in unquenched lattice QCD. We use the MILC highly improved staggered quark (HISQ) ensembles with four dynamical quark flavors. Our primary results are $f_{D^+} = 212.6(0.4)({}^{+1.0}_{-1.2})\ \mathrm{MeV}$, $f_{D_s} = 249.0(0.3)({}^{+1.1}_{-1.5})\ \mathrm{MeV}$, and $f_{D_s}/f_{D^+} = 1.1712(10)({}^{+29}_{-32})$, where the errors are statistical and total systematic, respectively. We also obtain $f_{K^+}/f_{\pi^+} = 1.1956(10)({}^{+26}_{-18})$, updating our previous result, and determine the quark-mass ratios $m_s/m_l = 27.35(5)({}^{+10}_{-7})$ and $m_c/m_s = 11.747(19)({}^{+59}_{-43})$. When combined with experimental measurements of the decay rates, our results lead to precise determinations of the CKM matrix elements $|V_{us}| = 0.22487(51) (29)(20)(5)$, $|V_{cd}|=0.217(1) (5)(1)$ and $|V_{cs}|= 1.010(5)(18)(6)$, where the errors are from this calculation of the decay constants, the uncertainty in the experimental decay rates, structure-dependent electromagnetic corrections, and, in the case of $|V_{us}|$, the uncertainty in $|V_{ud}|$, respectively.
This paper proposes a strategy to calibrate computational models according to uncertain input-output data. To this end, uncertainty in the data is first quantified by creating adversarial data sets. Samples drawn from such sets are then mapped from the input-output space to the parameter space using an inverse mapping. This mapping minimizes the collective output spread of an ensemble of point predictions while satisfying a set of individual data-matching requirements. The distribution of the resulting parameter points, which often exhibits strong parameter dependencies, is then modeled using sliced-normals. The chance-constrained formulation used to learn this distribution enables the analyst to trade-off a greater likelihood for most of the data against a lower likelihood for some of the data thereby relaxing the conservatism of the calibrated model. This formulation not only neglects the worst-performing quantiles of each adversarial distribution but also eliminates the potentially serious effects that outliers might have on the resulting model. This calibration approach not only has a considerably lower computational cost than the standard forward approach but it also allows for the identification of suitable distribution classes, which in turn yield better calibrated models.
In 1989, Sir Sam Edwards made the visionary proposition to treat jammed granular materials using a volume ensemble of equiprobable jammed states in analogy to thermal equilibrium statistical mechanics, despite their inherent athermal features. Since then, the statistical mechanics approach for jammed matter—one of the very few generalizations of Gibbs-Boltzmann statistical mechanics to out-of-equilibrium matter—has garnered an extraordinary amount of attention by both theorists and experimentalists. Its importance stems from the fact that jammed states of matter are ubiquitous in nature appearing in a broad range of granular and soft materials such as colloids, emulsions, glasses, and biomatter. Indeed, despite being one of the simplest states of matter—primarily governed by the steric interactions between the constitutive particles—a theoretical understanding based on first principles has proved exceedingly challenging. Here a systematic approach to jammed matter based on the Edwards statistical mechanical ensemble is reviewed. The construction of microcanonical and canonical ensembles based on the volume function, which replaces the Hamiltonian in jammed systems, is discussed. The importance of approximation schemes at various levels is emphasized leading to quantitative predictions for ensemble averaged quantities such as packing fractions and contact force distributions. An overview of the phenomenology of jammed states and experiments, simulations, and theoretical models scrutinizing the strong assumptions underlying Edwards approach is given including recent results suggesting the validity of Edwards ergodic hypothesis for jammed states. A theoretical framework for packings whose constitutive particles range from spherical to nonspherical shapes such as dimers, polymers, ellipsoids, spherocylinders or tetrahedra, hard and soft, frictional, frictionless and adhesive, monodisperse, and polydisperse particles in any dimensions is discussed providing insight into a unifying phase diagram for all jammed matter. Furthermore, the connection between the Edwards ensemble of metastable jammed states and metastability in spin glasses is established. This highlights the fact that the packing problem can be understood as a constraint satisfaction problem for excluded volume and force and torque balance leading to a unifying framework between the Edwards ensemble of equiprobable jammed states and out-of-equilibrium spin glasses.
Parallel training of a Deep Neural Network (DNN) ensemble on a cluster of nodes is a common practice to train multiple models in order to construct a model with a higher prediction accuracy. Existing ensemble training pipelines can perform a great deal of redundant operations, resulting in unnecessary CPU usage, or even poor pipeline performance. In order to remove these redundancies, we need pipelines with more communication flexibility than existing DNN frameworks can provide. This project investigates a series of designs to improve pipeline flexibility and adaptivity, while also increasing performance. We implement our designs using Tensorflow with Horovod, and test it using several large DNNs in a large scale GPU cluster, Titan supercomputer at Oak Ridge National Lab. Our results show that the CPU time spent during training is reduced by 2-11X. Furthermore, our implementation can achieve up to 10X speedups when CPU core limits are imposed. Our best pipeline also reduces the average power draw of the ensemble training process by 5-16% when compared to the baseline.
Ensemble is an open architecture for the development, integration, and deployment of mission operations software. Fundamentally, it is an adaptation of the Eclipse Rich Client Platform (RCP), a widespread, stable, and supported framework for component-based application development. By capitalizing on the maturity and availability of the Eclipse RCP, Ensemble offers a low-risk, politically neutral path towards a tighter integration of operations tools. The Ensemble project is a highly successful, ongoing collaboration among NASA Centers. Since 2004, the Ensemble project has supported the development of mission operations software for NASA's Exploration Systems, Science, and Space Operations Directorates.