SEARCH · Engineering Papers

Results for “NETWORK”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,567 records · Page 87

All sources NASA NTRS DOE OSTI

DOE OSTI2019

Towards Efficient Convolutional Neural Networks Through Low-Error Filter Saliency Estimation

Filter saliency based channel pruning is a state-of-the-art method for deep convolutional neural network compression and acceleration. This channel pruning method ranks the importance of individual filter by estimating its impact of each filter’s removal on the training loss, and then remove the least important filters and fine-tune the remnant network. In this work, we propose a systematic channel pruning method that significantly reduces the estimation error of filter saliency. Different from existing approaches, our method largely reduces the magnitude of parameters in a network by introducing alternating direction method of multipliers (ADMM) into the pre-training procedure. Therefore, the estimation of filter saliency based on Taylor expansion is significantly improved. Extensive experiments with various benchmark network architectures and datasets demonstrate that the proposed method has a much improved unimportant filter selection capability and outperform state-of-the-art channel pruning method.

Wang, Zi↗

DOE OSTI2018

Exploring the Role of Intrinsic Nodal Activation on the Spread of Influence in Complex Networks

In many complex networked systems such as online social networks, at any given time, activity originates at certain nodes and subsequently spreads on the network through influence. To model the spread of influence in such a scenario, we consider the problem of identification of influential entities in a complex network when nodal activation can happen through two different mechanisms. The first mode of activation is due mechanisms intrinsic to the node. The second mechanism is through the influence of connected neighbors. In this work, we present a simple probabilistic formulation that models such self-evolving systems where information diffusion occurs primarily because of the intrinsic activity of users and the spread of activity occurs due to influence. We provide an algorithm to mine for the influential seeds in such a scenario by modifying the well-known influence maximization framework with the independent cascade diffusion model. We provide small motivating examples to provide an intuitive understanding of the effect of including the intrinsic activation mechanism. We sketch a proof of the submodularity of the influence function under the new formulation and demonstrate the same with larger graphs. We then show by means of additional experiments on a real-world twitter dataset how the formulation can be applied to real-world social media datasets. Finally we derive a computationally efficient centrality metric that takes into account, both the mechanisms of activation and provides for an accurate as well as computationally efficient alternative approach to the problem of identifying influencers under intrinsic activation.

Visweswara Sathanur, Arun↗

DOE OSTI2019

Energy flow networks: deep sets for particle jets

A key question for machine learning approaches in particle physics is how to best represent and learn from collider events. As an event is intrinsically a variable-length unordered set of particles, we build upon recent machine learning efforts to learn directly from sets of features or “point clouds”. Adapting and specializing the “Deep Sets” framework to particle physics, we introduce Energy Flow Networks, which respect infrared and collinear safety by construction. We also develop Particle Flow Networks, which allow for general energy dependence and the inclusion of additional particle-level information such as charge and flavor. These networks feature a per-particle internal (latent) representation, and summing over all particles yields an overall event-level latent representation. We show how this latent space decomposition unifies existing event representations based on detector images and radiation moments. To demonstrate the power and simplicity of this set-based approach, we apply these networks to the collider task of discriminating quark jets from gluon jets, finding similar or improved performance compared to existing methods. We also show how the learned event representation can be directly visualized, providing insight into the inner workings of the model. These architectures lend themselves to efficiently processing and analyzing events for a wide variety of tasks at the Large Hadron Collider. Implementations and examples of our architectures are available online in our EnergyFlow package.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2021

Predicting Fracture Network Development in Crystalline Rocks

The geometric properties of fractures influence whether they propagate, arrest, or coalesce with other fractures. Thus, quantifying the relationship between fracture network characteristics may help predict fracture network development, and perhaps precursors to catastrophic failure. To constrain the relationship and predictability of fracture characteristics, we deform eight one centimeter tall rock cores under triaxial compression while acquiring in situ X-ray tomograms. The tomograms reveal precise measurements of the fracture network characteristics above the spatial resolution of 6.5 µm. We develop machine learning models to predict the value of each characteristic using the other characteristics, and excluding the macroscopic stress or strain imposed on the rock. The models predict fracture development more accurately in the experiments performed on granite and monzonite, than the experiments on marble. Fracture network development may be more predictable in these igneous rocks because their microstructure is more mechanically homogeneous than the marble, producing more systematic fracture development that is not strongly impeded by grain contacts and cleavage planes. The varying performance of the models suggest that fracture volume, length, and aperture are the most predictable of the characteristics, while fracture orientation is the least predictable. Orientation does not correlate with length, as suggested by the idea that the orientation evolves with increasing differential stress and thus fracture length. This difference between the observed and expected relationship between orientation and length highlights the influence of mechanical heterogeneities and local stress perturbations on fracture growth as fractures propagate, link, and coalesce.

58 GEOSCIENCES↗

DOE OSTI2020

Detecting anomalous packets in network transfers: investigations using PCA, autoencoder and isolation forest in TCP

Large-scale scientific workflows rely heavily on high-performance file transfers. These transfers require strict quality parameters such as guaranteed bandwidth, no packet loss or data duplication. To have successful file transfers, methods such as predetermined thresholds and statistical analysis need to be done to determine abnormal patterns. Network administrators routinely monitor and analyze network data for diagnosing and alleviating these, making decisions based on their experience. However, as networks grow and become complex, monitoring large data files and quickly processing them, makes it improbable to identify errors and rectify these. Abnormal file transfers have been classified by simply setting alert thresholds, via tools such as PerfSonar and TCP statistics (Tstat). This paper investigates the feasibility of unsupervised feature extraction methods for identifying network anomaly patterns with three unsupervised classification methods—principal component analysis, autoencoder and isolation forest. Here, we collect file transfer statistics from two experiment sets—synthetic iPerf generated traffic and 1000 Genome workflow runs, with synthetically introduced anomalies. Our results show that while PCA and a simple autoencoder finds it difficult to detect clusters, the tree-variant isolation forest is able to identify anomalous packets by breaking down TCP traces into tree classes early.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2016

Determining the Impact of Personal Mobility Carbon Allowance Schemes in Transportation Networks

We know that personal mobility carbon allowance (PMCA) schemes are designed to reduce carbon consumption from transportation networks. PMCA schemes influence the travel decision process of users and accordingly impact the system metrics including travel time and greenhouse gas (GHG) emissions. Here, we develop a multi-user class dynamic user equilibrium model to evaluate the transportation system performance when PMCA scheme is implemented. The results using Sioux-Falls test network indicate that PMCA schemes can achieve the emissions reduction goals for transportation networks. Further, users characterized by high value of travel time are found to be less sensitive to carbon budget in the context of work trips. Results also show that PMCA scheme can lead to higher emissions for a path compared with the case without PMCA because of flow redistribution. The developed network equilibrium model allows us to examine the change in system states at different carbon allocation levels and to design parameters of PMCA schemes accounting for population heterogeneity.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

Scale-bridging Optimization Framework for Desalination Integrated Produced Water Networks

Produced water (PW) management poses a major challenge to U.S. oil and gas development. Due to freshwater scarcity and PW injection curtailments, desalination may become a necessity in some parts of the country. Integrating rigorous desalination process models within a produced water network for co-optimization of desalination design and network operation can be challenging to scale-up due to computational complexity. In this work we use the trust region filter framework to decompose the integrated optimization problem into an optimization master problem consisting of the network model and simplified surrogates for the desalination units and an optimization sub problem for the rigorous desalination unit model. The approach is demonstrated on a multiperiod produced water network from the PARETO library with thermal desalination units.

Naik, Sakshi↗

DOE OSTI2016

Three-dimensional dendritic needle network model for alloy solidification

Here, we present a three-dimensional (3D) formulation of the multiscale Dendritic Needle Network (DNN) model for dendritic microstructure growth. This approach is aimed at simulating quantitatively the solidification dynamics of complex hierarchical networks in spatially extended dendritic arrays, hence bridging the scale gap between phase-field simulations at the scale of a few dendrites and coarse-grained simulations on the larger scale of entire polycrystalline structures. In the DNN model, the dendritic network is represented by a network of branches that interact through the solutal diffusion field. The tip velocity V(t) and tip radius ρ(t) of each needle is determined by combining a standard solvability condition that fixes the product ρ 2 V and a solute flux conservation condition that fixes the product ρV 2 in 2D and ρV in 3D as a function of a solutal flux intensity factor ℱ(t). The latter measures the intensity of the solute flux in the dendrite tip region and can be calculated by contour (2D) or surface (3D) integration around the tip of each needle. We first present an extended formulation of the 2D DNN model where needles have a finite thickness and parabolic tips. This formulation remains valid for a larger range of tip Péclet number than the original thin needle formulation and is readily extended to 3D needles with paraboloidal tips. The 3D DNN model based on this thick-needle formulation is developed for both isothermal and directional solidification. Model predictions are validated by comparisons with known analytical solutions that describe the early transient and steady-state growth regimes. We exploit the power of the DNN model to characterize the competitive growth of well-developed secondary branches in 3D on the scale of the diffusion length. The results show that the length of active secondary branches increases as a power law of distance behind the tip with an exponent in good quantitative agreement with experimental measurements. Finally, we apply the model to simulate the three-dimensional directional solidification of an Al-7wt% Si alloy, which we directly compare to observed microstructures from microgravity experiments onboard the International Space Station. The predictions of selected microstructural features, such as dendrite arm spacings, show a good agreement with experiments. The computationally-efficient DNN model opens new avenues for investigating the dynamics of large dendritic arrays at length and time scales relevant to solidification experiments and processes.

36 MATERIALS SCIENCE↗

DOE OSTI2012

Non-abelian symmetries in tensor networks: A quantum symmetry space approach

A general framework for non-abelian symmetries is presented for matrix-product and tensor-network states in the presence of well-defined orthonormal local as well as effective basis sets. The two crucial ingredients, the Clebsch-Gordan algebra for multiplet spaces as well as the Wigner-Eckart theorem for operators, are accounted for in a natural, well-organized, and computationally straightforward way. The unifying tensor-representation for quantum symmetry spaces, dubbed QSpace, is particularly suitable to deal with standard renormalization group algorithms such as the numerical renormalization group (NRG), the density matrix renormalization group (DMRG), or also more general tensor networks such as the multi-scale entanglement renormalization ansatz (MERA). In this paper, the focus is on the application of the non-abelian framework within the NRG. A detailed analysis is presented for a fully screened spin- 3/2 three-channel Anderson impurity model in the presence of conservation of total spin, particle-hole symmetry, and SU(3) channel symmetry. The same system is analyzed using several alternative symmetry scenarios based on combinations of U(1){sub charge}, SU(2){sub spin}, SU(2){sub charge}, SU(3){sub channel}, as well as the enveloping symplectic Sp(6) symmetry. These are compared in detail, including their respective dramatic gain in numerical efficiency. In the Appendix, finally, an extensive introduction to non-abelian symmetries is given for practical applications, together with simple self-contained numerical procedures to obtain Clebsch-Gordan coefficients and irreducible operators sets. The resulting QSpace tensors can deal with any set of abelian symmetries together with arbitrary non-abelian symmetries with compact, i.e. finite-dimensional, semi-simple Lie algebras. - Highlights: Black-Right-Pointing-Pointer We introduce a transparent framework for non-abelian symmetries in tensor networks. Black-Right-Pointing-Pointer The framework was successfully applied within the numerical renormalization group. Black-Right-Pointing-Pointer Combinations of SU(N{<=}3) and Sp(6) are implemented and discussed in detail. Black-Right-Pointing-Pointer Non-abelian symmetries allow to speed up calculations by orders of magnitude. Black-Right-Pointing-Pointer The framework is equally applicable to other tensor networks such as DMRG or MERA.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2014

A practical introduction to tensor networks: Matrix product states and projected entangled pair states

This is a partly non-technical introduction to selected topics on tensor network methods, based on several lectures and introductory seminars given on the subject. It should be a good place for newcomers to get familiarized with some of the key ideas in the field, specially regarding the numerics. After a very general introduction we motivate the concept of tensor network and provide several examples. We then move on to explain some basics about Matrix Product States (MPS) and Projected Entangled Pair States (PEPS). Selected details on some of the associated numerical methods for 1d and 2d quantum lattice systems are also discussed. - Highlights: • A practical introduction to selected aspects of tensor network methods is presented. • We provide analytical examples of MPS and 2d PEPS. • We provide basic aspects on several numerical methods for MPS and 2d PEPS. • We discuss a number of applications of tensor network methods from a broad perspective.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2024

Simulation of gas mixture dynamics in a pipeline network using explicit staggered-grid discretization

Here we develop an explicit staggered finite difference discretization scheme for simulating the transport of highly heterogeneous gas mixtures through pipeline networks. This study is motivated by the proposed blending of hydrogen into natural gas pipelines to reduce end use carbon emissions while using existing pipeline systems throughout their planned lifetimes. Our computational method accommodates an arbitrary number of constituent gases with very different physical properties that may be injected into a network with significant spatiotemporal variation. In this setting, the gas flow physics are highly location- and time- dependent, so that local composition and nodal mixing must be accounted for. The resulting conservation laws are formulated in terms of pressure, partial densities and flows, and volumetric and mass fractions of the constituents. We include non-ideal equations of state that employ linear approximations of gas compressibility factors, so that the pressure dynamics propagate locally according to a variable wave speed that depends on mixture composition and density. We derive compatibility relationships for network edge boundary values that are more complex than for a homogeneous gas. The simulation method is evaluated on initial boundary value problems for a single pipe and a small network, is cross-validated with a lumped element simulation, and used to demonstrate a local monitoring and control policy for maintaining allowable concentration levels.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2019

Classifying Cancer Pathology Reports with Hierarchical Self-Attention Networks

We introduce a deep learning architecture, hierarchical self-attention networks (HiSANs), designed for classifying pathology reports and show how its unique architecture leads to a new state-of-the-art in accuracy, faster training, and clear interpretability. We evaluate performance on a corpus of 374,899 pathology reports obtained from the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) program. Each pathology report is associated with five clinical classification tasks – site, laterality, behavior, histology, and grade. We compare the performance of the HiSAN against other machine learning and deep learning approaches commonly used on medical text data – Naive Bayes, logistic regression, convolutional neural networks, and hierarchical attention networks (the previous state-of-the-art). We show that HiSANs are superior to other machine learning and deep learning text classifiers in both accuracy and macro F-score across all five classification tasks. Compared to the previous state-of-the-art, hierarchical attention networks, HiSANs not only are an order of magnitude faster to train, but also achieve about 1% better relative accuracy and 5% better relative macro F-score.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2015

Transcription factor FOXA2-centered transcriptional regulation network in non-small cell lung cancer

Lung cancer is the leading cause of cancer-mediated death. Although various therapeutic approaches are used for lung cancer treatment, these mainly target the tumor suppressor p53 transcription factor, which is involved in apoptosis and cell cycle arrest. However, p53-targeted therapies have limited application in lung cancer, since p53 is found to be mutated in more than half of lung cancers. In this study, we propose tumor suppressor FOXA2 as an alternative target protein for therapies against lung cancer and reveal a possible FOXA2-centered transcriptional regulation network by identifying new target genes and binding partners of FOXA2 by using various screening techniques. The genes encoding Glu/Asp-rich carboxy-terminal domain 2 (CITED2), nuclear receptor subfamily 0, group B, member 2 (NR0B2), cell adhesion molecule 1 (CADM1) and BCL2-associated X protein (BAX) were identified as putative target genes of FOXA2. Additionally, the proteins including highly similar to heat shock protein HSP 90-beta (HSP90A), heat shock 70 kDa protein 1A variant (HSPA1A), histone deacetylase 1 (HDAC1) and HDAC3 were identified as novel interacting partners of FOXA2. Moreover, we showed that FOXA2-dependent promoter activation of BAX and p21 genes is significantly reduced via physical interactions between the identified binding partners and FOXA2. These results provide opportunities to understand the FOXA2-centered transcriptional regulation network and novel therapeutic targets to modulate this network in p53-deficient lung cancer. - Highlights: • Identification of new target genes of FOXA2. • Identifications of novel interaction proteins of FOXA2. • Construction of FOXA2-centered transcriptional regulatory network in non-small cell lung cancer.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2018

Uncovering transcription factor and microRNA risk regulatory pathways associated with osteoarthritis by network analysis

Osteoarthritis (OA) is the most common form of joint disease. The development of inflammation have been considered to play a key role during the progression of OA. Regulatory pathways are known to play crucial roles in many pathogenic processes. Thus, deciphering these risk regulatory pathways is critical for elucidating the mechanisms underlying OA. We constructed an OA-specific regulatory network by integrating comprehensive curated transcription and post-transcriptional resource involving transcription factor (TF) and microRNA (miRNA). To deepen our understanding of underlying molecular mechanisms of OA, we developed an integrated systems approach to identify OA-specific risk regulatory pathways. In this study, we identified 89 significantly differentially expressed genes between normal and inflamed areas of OA patients. We found the OA-specific regulatory network was a standard scale-free network with small-world properties. It significant enriched many immune response-related functions including leukocyte differentiation, myeloid differentiation and T cell activation. Finally, 141 risk regulatory pathways were identified based on OA-specific regulatory network, which contains some known regulator of OA. The risk regulatory pathways may provide clues for the etiology of OA and be a potential resource for the discovery of novel OA-associated disease genes.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2024

Extreme sparsification of physics-augmented neural networks for interpretable model discovery in mechanics

Data-driven constitutive modeling with neural networks has received increased interest in recent years due to its ability to easily incorporate physical and mechanistic constraints and to overcome the challenging and time-consuming task of formulating phenomenological constitutive laws that can accurately capture the observed material response. However, even though neural network-based constitutive laws have been shown to generalize proficiently, the generated representations are not easily interpretable due to their high number of trainable parameters. Sparse regression approaches exist that allow for obtaining interpretable expressions, but the user is tasked with creating a library of model forms which by construction limits their expressiveness to the functional forms provided in the libraries. Here, in this work, we propose to train regularized physics-augmented neural network-based constitutive models utilizing a smoothed version of $L^0$-regularization. This aims to maintain the trustworthiness inherited by the physical constraints, but also enables interpretability which has not been possible thus far on any type of machine learning-based constitutive model where model forms were not assumed a priori but were actually discovered. During the training process, the network simultaneously fits the training data and penalizes the number of active parameters, while also ensuring constitutive constraints such as thermodynamic consistency. We show that the method can reliably obtain interpretable and trustworthy constitutive models for compressible and incompressible hyperelasticity, yield functions, and hardening models for elastoplasticity, using synthetic and experimental data. This work aims to set a new paradigm for interpretable machine learning models in the broad area of solid mechanics where low and limited data is available along with prior knowledge of physical constraints that the learned maps need to obey. This paradigm can potentially be extended to a broader spectrum of scientific exploration.

Data-driven constitutive models↗

DOE OSTI2025

Constrained or unconstrained? Neural-network-based equation discovery from data

Throughout many fields, practitioners often rely on differential equations to model systems. Yet, for many applications, the theoretical derivation of such equations and/or the accurate resolution of their solutions may be intractable. Instead, recently developed methods, including those based on parameter estimation, operator subset selection, and neural networks, allow for the data-driven discovery of both ordinary and partial differential equations (PDEs), on a spectrum of interpretability. The success of these strategies is often contingent upon the correct identification of representative equations from noisy observations of state variables and, as importantly and intertwined with that, the mathematical strategies utilized to enforce those equations. Specifically, the latter has been commonly addressed via unconstrained optimization strategies. Representing the PDE as a neural network, we propose to discover the PDE (or the associated operator) by solving a constrained optimization problem and using an intermediate state representation similar to a physics-informed neural network (PINN). The objective function of this constrained optimization problem promotes matching the data, while the constraints require that the discovered PDE is satisfied at a number of spatial collocation points. We present a penalty method and a widely used trust-region barrier method to solve this constrained optimization problem, and we compare these methods on numerical examples. Our results on several example problems demonstrate that the latter constrained method outperforms the penalty method, particularly for higher noise levels or fewer collocation points. This work motivates further exploration into using sophisticated constrained optimization methods in scientific machine learning, as opposed to their commonly used, penalty-method or unconstrained counterparts. For both of these methods, we solve these discovered neural network PDEs with classical methods, such as finite difference methods, as opposed to PINNs-type methods relying on automatic differentiation. Here, we briefly highlight how simultaneously fitting the data while discovering the PDE improves the robustness to noise and other small, yet crucial, implementation details.

Data-driven discovery↗

DOE OSTI2017

Privacy-aware contextual localization using network traffic analysis

The rise of location-based services has enabled many opportunities for content service providers to optimize the content delivery to user's wireless devices based on her location. Since the sharing precise location remains a major privacy concern among the users, certain location-based services rely on contextual location (e.g. residence, work, etc.) as opposed to acquiring user's exact physical location. In this paper, we present PACL (Privacy-Aware Contextual Localizer) model, which can learn user's contextual location just by passively monitoring user's network traffic. PACL can discern a set of vital attributes (statistical and application-based) from user's network traffic, and predict user's contextual location with a very high accuracy. We design and evaluate PACL using real-world network traces of over 1700 users with over 100GB of total data. Our results show that PACL, when built using the Bayesian Network machine learning algorithm, can predict user's contextual location with the accuracy of around 89%.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

Geometrical and chemical effects on the electrochemistry of single-wall carbon nanotube (SWCNT) network electrodes

Single-wall carbon nanotube (SWCNT) network is a promising electrode material for bio detection. Unfortunately, the associations between their physical as well as chemical properties and observed electrochemical performance are not known. This hinders any systematic optimization of the network properties towards specific analytes. Here we present a consistent physicochemical and electrochemical characterization of differently treated SWCNT networks. The results unambiguously show that (i) even if the electrochemical properties of different electrodes are practically identical when assessed by surface insensitive outer sphere redox (OSR) probes their behavior with inner sphere redox (ISR) probes can be drastically different. Further, (ii) the choice of the modification method (structural, chemical, electrochemical) heavily depends on nature of the target analyte, which are typically ISR probes. Although, (iii) chemical changes in the carbon phase appeared to be minor, effects of different treatments on oxidation states of Fe appeared to have a strong effect on the electrochemical performance of the networks in the case of ISR probes.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗