Search NASA⌕ Search

SEARCH · Search NASA

Results for “Network Performance”

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 433 records · Page 24

Characterizing 4-string contact interaction using machine learning

Abstract The geometry of 4-string contact interaction of closed string field theory is characterized using machine learning. We obtain Strebel quadratic differentials on 4-punctured spheres as a neural network by performing unsupervised learning with a custom-built loss function. This allows us to solve for local coordinates and compute their associated mapping radii numerically. We also train a neural network distinguishing vertex from Feynman region. As a check, 4-tachyon contact term in the tachyon potential is computed and a good agreement with the results in the literature is observed. We argue that our algorithm is manifestly independent of number of punctures and scaling it to characterize the geometry ofn-string contact interaction is feasible.

Physics↗

Macroscopic Traffic Modeling Using Probe Vehicle Data: A Machine Learning Approach

Abstract The macroscopic fundamental diagram (MFD) captures an orderly relationship among traffic flow, density, and speed at the network level. It is a simple yet powerful tool for modeling traffic dynamics in large urban networks with broad application in traffic control and management. However, empirically derived MFDs in urban regions require high-resolution traffic data from the network. Having the network flow and vehicular density estimated at the (granular) census tract level using vehicle probe data, we apply machine learning methods to predict the MFDs across U.S. urban areas and capture the impacts of location-specific input features on the network flow–density relationships at a large scale. The results show that, among the four tested machine learning approaches (Random Forest, XGBoost, Support Vector Machine, and Neural Network), XGBoost delivers the best performance in predicting network traffic flow based on vehicular density and location attributes. Using interaction Shapley Additive explanation (SHAP) values and partial correlation analysis, we examine the factors influencing MFD shapes across different locations. Our empirical findings reveal that across U.S. urban areas, network topology, transportation infrastructure, and land use are primary factors shaping MFD curves, while demand and trip-related factors play a lesser role. Specifically, higher ranking roads, centrality, and development levels correlate positively with network capacity and critical density, whereas negative associations are observed for network connectivity, mixed-use development, and road roughness levels.

Jin, Ling↗

BSTC: A Novel Binarized-Soft-Tensor-Core Design for Accelerating Bit-Based Approximated Neural Nets

Binarized neural network (or BNN) promises tremendous performance improvement over traditional DNNs through simplified bit-level computation and significantly reduced memory access/storage cost. In addition, it has other advantages of low-cost, low-energy and high-robustness, showing great utilization potential in resources-constrained, volatile and latency-critical applications, which are critical for future HPC execution. However, the promised significant performance gain of BNN inference has never been fully demonstrated on general-purpose processors, particularly on GPUs, due to: (i) the challenge to extract and leverage sufficient fine-grained bit-level-parallelism to saturate GPU cores when batch is small; (ii) the fundamental design conflict between bit-based BNN algorithm and underlying word-based architecture; (iii) architecture \& performance unfriendly BNN network design. To address (i) and (ii), we propose binarized-soft-tensor-core as a software-hardware codesign approach to construct bit-manipulation capability for modern GPUs to effectively harvest the emerging bit-level-parallelism. To tackle (iii), we propose intra- and inter-layer fusion techniques so that the entire BNN inference can be packed into a single GPU kernel, to avoid high-cost frequent launching. Experiments demonstrate that our design can achieve over 1000x speedup for raw inference latency and 10x for inference throughput over the state-of-the-art full-precision simulated BNN inference for AlexNet on ImageNet.

Li, Ang↗

Mass measurements of neutron rich isotopes in the Fe region and electron capture processes in neutron star crusts

Experimental knowledge of nuclear masses of exotic nuclei is important for understanding nu- clear structure far from the valley of -stability, and as a direct input into astrophysical models. Electron capture processes in the crust of accreting neutron stars have been proposed as a heat source that can affect the thermal structure of the star. Nuclear masses of very neutron-rich nu- clides are necessary inputs to model the electron capture process. The time-of-flight (TOF) mass measurement technique allows measurements on very short-lived nuclei. It has been effectively applied using the fast fragment beams produced at the National Superconducting Cyclotron Lab (NSCL) to reach masses very far from stability. Measurements were performed for neutron-rich isotopes in the region of the N=32 and N=40 subshells, which coincides with the mass range of carbon superburst ashes. We discuss reaction network calculations performed to investigate the impact of our new measurements and to compare the effect of using different global mass models in the calculations. It is observed that the process is sensitive to the differences in the odd-even mass staggering predicted by the mass models, and our new result for 66Mn has a significant impact on the distribution of heat sources in the crust.

43 PARTICLE ACCELERATORS↗

Laser Tracker Calibration - Testing the Angle Measurement System -

Physics experiments at the SLAC National Accelerator Laboratory (SLAC) usually require high accuracy positioning, e. g. 100 {micro}m over a distance of 150 m or 25 {micro}m in a 10 x 10 x 3 meter volume. Laser tracker measurement systems have become one of the most important tools for achieving these accuracies when mapping components. The accuracy of these measurements is related to the manufacturing tolerances of various individual components, the resolutions of measurement systems, the overall precision of the assembly, and how well imperfections can be modeled. As with theodolites and total stations, one can remove the effects of most assembly and calibration errors by measuring targets in both direct and reverse positions and computing the mean to obtain the result. However, this approach does not compensate for errors originating from the encoder system. In order to improve and gain a better understanding of laser tracker angle measurement tolerances we extended our laboratory's capabilities with the addition of a horizontal angle calibration test stand. This setup is based on the use of a high precision rotary table providing an angular accuracy of better than 0.2 arcsec. Presently, our setup permits only tests of the horizontal angle measurement system. A test stand for vertical angle calibration is under construction. Distance measurements (LECOCQ & FUSS, 2000) are compared to an interferometer bench for distances of up to 32 m. Together both tests provide a better understanding of the instrument and how it should be operated. The observations also provide a reasonable estimate of covariance information of the measurements according to their actual performance for network adjustments.

43 PARTICLE ACCELERATORS↗

Autoencoder node saliency: Selecting relevant latent representations

The autoencoder is an artificial neural network that performs nonlinear dimension reduction and learns hidden representations of unlabeled data. With a linear transfer function it is similar to the principal component analysis (PCA). While both methods use weight vectors for linear transformations, the autoencoder does not come with any indication similar to the eigenvalues in PCA that are paired with eigenvectors. In this work, we propose a novel autoencoder node saliency method that examines whether the features constructed by autoencoders exhibit properties related to known class labels. The supervised node saliency ranks the nodes based on their capability of performing a learning task. It is coupled with the normalized entropy difference (NED). We establish a property for NED values to verify classifying behaviors among the top ranked nodes. Lastly, by applying our methods to real datasets, we demonstrate their ability to provide indications on the performing nodes and explain the learned tasks in autoencoders.

97 MATHEMATICS AND COMPUTING↗

Combining Spike Time Dependent Plasticity (STDP) and Backpropagation (BP) for Robust and Data Efficient Spiking Neural Networks (SNN)

National security applications require artificial neural networks (ANNs) that consume less power, are fast and dynamic online learners, are fault tolerant, and can learn from unlabeled and imbalanced data. We explore whether two fundamentally different, traditional learning algorithms from artificial intelligence and the biological brain can be merged. We tackle this problem from two directions. First, we start from a theoretical point of view and show that the spike time dependent plasticity (STDP) learning curve observed in biological networks can be derived using the mathematical framework of backpropagation through time. Second, we show that transmission delays, as observed in biological networks, improve the ability of spiking networks to perform classification when trained using a backpropagation of error (BP) method. These results provide evidence that STDP could be compatible with a BP learning rule. Combining these learning algorithms will likely lead to networks more capable of meeting our national security missions.

97 MATHEMATICS AND COMPUTING↗

Demonstration of Decentralized Physics-Driven Learning

In typical artificial neural networks, neurons adjust according to global calculations of a central processor, but in the brain, neurons and synapses self-adjust based on local information. Contrastive learning algorithms have recently been proposed to train physical systems, such as fluidic, mechanical, or electrical networks, to perform machine-learning tasks from local evolution rules. However, to date, such systems have only been implemented in silico due to the engineering challenge of creating elements that autonomously evolve based on their own response to two sets of global boundary conditions. Here, we introduce and implement a physics-driven contrastive learning scheme for a network of variable resistors, using circuitry to locally compare the response of two identical networks subjected to the two different sets of boundary conditions. Using this method, our system effectively trains itself, optimizing its resistance values without the use of a central processor or external information storage. Once the system is trained for a specified allostery, regression, or classification task, the task is subsequently performed rapidly and automatically by the physical imperative to minimize power dissipation in response to the given voltage inputs. Here, we demonstrate that, unlike typical computers, such learning systems are robust to extreme damage (and thus manufacturing defects) due to their decentralized learning. Our twin-network approach is therefore readily scalable to extremely large or nonlinear networks, where its distributed nature will be an enormous advantage; a laboratory network of only 500 edges will already outpace its in silico counterpart.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Chaotic neural dynamics facilitate probabilistic computations through sampling

Cortical neurons exhibit highly variable responses over trials and time. Theoretical works posit that this variability arises potentially from chaotic network dynamics of recurrently connected neurons. Here, we demonstrate that chaotic neural dynamics, formed through synaptic learning, allow networks to perform sensory cue integration in a sampling-based implementation. We show that the emergent chaotic dynamics provide neural substrates for generating samples not only of a static variable but also of a dynamical trajectory, where generic recurrent networks acquire these abilities with a biologically plausible learning rule through trial and error. Furthermore, the networks generalize their experience in the stimulus-evoked samples to the inference without partial or all sensory information, which suggests a computational role of spontaneous activity as a representation of the priors as well as a tractable biological computation for marginal distributions. These findings suggest that chaotic neural dynamics may serve for the brain function as a Bayesian generative model.

60 APPLIED LIFE SCIENCES↗

Eulerian network modeling of longitudinal dispersion: EULERIAN NETWORK MODELING OF DISPERSION

A novel Eulerian network model that incorporates “shear dispersion,” the stretching of solute due to nonuniform velocity profiles within pore throats is developed. The superposing transport method (STM) is nonlocal in time (i.e., uses information from several previous time steps) and is equivalent to performing network-wide time convolutions of elementary throat response functions. Predicted macroscopic longitudinal dispersion coefficients for disordered sphere packs are in good agreement with published experimental data. We further investigate the impact of mixing assumptions within pores on macroscopic longitudinal dispersion and find the dependence to be weak for disordered sphere packs. Limitations of Eulerian network models as a whole are also discussed, and their inappropriateness for ordered porous media concluded.

58 GEOSCIENCES↗

Effect of Strong Opinions on the Dynamics of the Majority-Vote Model

We study how the presence of individuals with strong opinions affects a square lattice majority-vote model with noise. In a square lattice network we perform Monte-Carlo simulations and replace regular actors σ with strong actors μ in a random distribution. We find that the value of the critical noise parameter q c is a decreasing function of the concentration r of strong actors in the social interaction network. We calculate the critical exponents β/ν, γ/ν, and 1/ν and find that the presence of strong actors does not change the Ising universality class of the isotropic majority-vote model.

97 MATHEMATICS AND COMPUTING↗

Inferring Convolutional Neural Networks' Accuracies from Their Architectural Characterizations

The challenge of choosing an appropriate convolutional neural network (CNN) architecture for specific applications and different data sets is still poorly understood in the literature. This is problematic, since CNNs have shown strong promise for analyzing scientific data from many domains including particle imaging detectors. In this paper, we proposed a systematic language that is useful for comparison between different CNN's architectures before training time. This helps us predict whether a network can perform better than a certain threshold accuracy before training up to 70% accuracy using simple machine learning models. Additionally, we found a coefficient of determination of 0.966 for an Ordinary Least Squares model in a regression task to predict accuracy of a large population of networks.

Hoang, Duc↗

DART 7.0 User Guide

The Desktop Analysis Reporting Tool (DART) is a software package that allows users to easily view and analyze daily files that span long periods. DART gives users the capability to quickly determine the state of health of a radiation portal monitor (RPM), troubleshoot and diagnose problems, and view data in various time frames to perform trend analysis. In short, it converts the data strings written in the daily files into meaningful tables and plots. The standalone version of DART (“soloDART”) utilizes a database engine that is included with the application; no additional installations are necessary. There is also a networked version of DART (“polyDART”) that is designed to maximize the benefit of a centralized data repository while distributing the workload to individual desktop machines. This networked approach requires a more complex database manager Structured Query Language (SQL) Server; however, SQL Server is not currently provided with DART. Regardless of which version is used, DART will import daily files from RPMs, store the relevant data in its database, and it can produce reports for status, trend analysis, and reporting purposes.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

A game theoretic investigation of deception in network security

We perform a game theoretic investigation of the effects of deception on the interactions between an attacker and a defender of a computer network. The defender can employ camouflage by either disguising a normal system as a honeypot or by disguising a honeypot as a normal system. We model the interactions between defender and attacker using a signaling game, a non-cooperative two player dynamic game of incomplete information. For this model, we determine which strategies admit perfect Bayesian equilibria. These equilibria are refined Nash equilibria in which neither the defender nor the attacker will unilaterally choose to deviate from their strategies. Finally, we discuss the benefits of employing deceptive equilibrium strategies in the defense of a computer network.

99 GENERAL AND MISCELLANEOUS↗

Quantifying uncertainty for deep learning based forecasting and flow-reconstruction using neural architecture search ensembles

Classical problems in computational physics such as data-driven forecasting and signal reconstruction from sparse sensors have recently seen an explosion in deep neural network (DNN) based algorithmic approaches. However, most DNN models do not provide uncertainty estimates, which are crucial for establishing the trustworthiness of these techniques in downstream decision making tasks and scenarios. In recent years, ensemble-based methods have achieved significant success for the uncertainty quantification in DNNs on a number of benchmark problems. However, their performance on real-world applications remains under-explored. In this work, we present an automated approach to DNN discovery and demonstrate how this may also be utilized for ensemble-based uncertainty quantification. Specifically, we propose the use of a scalable neural and hyperparameter architecture search for discovering an ensemble of DNN models for complex dynamical systems. We highlight how the proposed method not only discovers high-performing neural network ensembles for our tasks, but also quantifies uncertainty seamlessly. This is achieved by using genetic algorithms and Bayesian optimization for sampling the search space of neural network architectures and hyperparameters. Subsequently, a model selection approach is used to identify candidate models for an ensemble set construction. Afterwards, a variance decomposition approach is used to estimate the uncertainty of the predictions from the ensemble. We demonstrate the feasibility of this framework for two tasks — forecasting from historical data and flow reconstruction from sparse sensors for the sea-surface temperature. In conclusion, we demonstrate superior performance from the ensemble in contrast with individual high-performing models and other benchmarks.

Deep ensembles↗

Verification of the astrometric performance of the Korean VLBI network, using comparative SFPR studies with the VLBA AT 14/7 mm

The Korean VLBI Network (KVN) is a new millimeter VLBI dedicated array with the capability to simultaneously observe at multiple frequencies, up to 129 GHz. The innovative multi-channel receivers present significant benefits for astrometric measurements in the frequency domain. The aim of this work is to verify the astrometric performance of the KVN using a comparative study with the VLBA, a well-established instrument. For that purpose, we carried out nearly contemporaneous observations with the KVN and the VLBA, at 14/7 mm, in 2013 April. The KVN observations consisted of simultaneous dual frequency observations, while the VLBA used fast frequency switching observations. We used the Source Frequency Phase Referencing technique for the observational and analysis strategy. We find that having simultaneous observations results in superior compensation for all atmospheric terms in the observables, in addition to offering other significant benefits for astrometric analysis. We have compared the KVN astrometry measurements to those from the VLBA. We find that the structure blending effects introduce dominant systematic astrometric shifts, and these need to be taken into account. We have tested multiple analytical routes to characterize the impact of the low-resolution effects for extended sources in the astrometric measurements. The results from the analysis of the KVN and full VLBA data sets agree within 2σ of the thermal error estimate. We interpret the discrepancy as arising from the different resolutions. We find that the KVN provides astrometric results with excellent agreement, within 1σ, when compared to a VLBA configuration that has a similar resolution. Therefore, this comparative study verifies the astrometric performance of the KVN using SFPR at 14/7 mm, and validates the KVN as an astrometric instrument.

79 ASTRONOMY AND ASTROPHYSICS↗