Search NASA⌕ Search

SEARCH · Search NASA

Results for “Complex Network Data”

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 1,243 records · Page 69

DOE BSSD Performance Management Metrics Report Q1

Microbes play key roles in our biosphere, from driving global nutrient cycling to impacting plant, animal and human health and disease. Complex data from microbial genomes, proteins, and metabolites provide a window into these tiny engines that drive life on our planet. Yet these data are dispersed among researchers’ laboratories and various repositories, making it difficult to access. This calls for new ways of managing data, improving data interoperability, advancing community standards, and creating an infrastructure where data are shared efficiently. We have built the National Microbiome Data Collaborative (NMDC) to advance how scientists create, use, and reuse data to redefine the way we understand and harness the power of microbes. The vision of the National Microbiome Data Collaborative (NMDC) is to drive a microbiome data sharing network connecting data, people, and ideas to advance microbiome innovation and discovery. The NMDC was launched in 2019 and brought together DOE National Laboratories to collaborate across resources, capabilities, and expertise. The NMDC team was strategically assembled to include software developers, microbial researchers, metadata experts, and multi-omics specialists. The diversity of the NMDC team reflects the inherently interdisciplinary nature of microbiome science, and we leverage the strengths of the DOE National Laboratory system. Towards BER’s goal of advancing an iterative systems biology approach to the understanding of microbial genomes, the NMDC serves as a foundation for infrastructure, data standards, and community building. Together with the flagship DOE User Facilities, the Joint Genome Institute (JGI) and the Environmental Molecular Sciences Laboratory (EMSL), we are developing core capabilities in metadata standards for environmental descriptors and sample handling and processing; standardized bioinformatic workflows; an interface for data search and access; and robust community engagement activities. The NMDC production platform supports long-term data infrastructure and community building for BER’s bioenergy and environmental research goals. Our approach leverages lessons learned and an ambitious framework for collaborative, interdisciplinary data infrastructure to support microbiome research. The NMDC supports data, information, and knowledge access through three defined software tools – the Submission Portal, NMDC EDGE, and the Data Portal – driven by community needs. Herein, we describe the value proposition for the microbiome research community, our overarching strategy, and challenges and opportunities for developing the NMDC as both an infrastructure and community engagement program.

59 BASIC BIOLOGICAL SCIENCES↗

Performance Prediction of Big Data Transfer Through Experimental Analysis and Machine Learning

Big data transfer in next-generation scientific ap- plications is now commonly carried out over connections with guaranteed bandwidth provisioned in High-performance Net- works (HPNs) through advance bandwidth reservation. To use HPN resources efficiently, provisioning agents need to carefully schedule data transfer requests and allocate appropriate band- widths. Such reserved bandwidths, if not fully utilized by the requesting user, could be simply wasted or cause extra overhead and complexity in management due to exclusive access. This calls for the capability of performance prediction to reserve bandwidth resources that match actual needs. Towards this goal, we employ machine learning algorithms to predict big data transfer perfor- mance based on extensive performance measurements, which are collected over a span of several years from a large number of data transfer tests using different protocols and toolkits between various end sites on several real-life physical or emulated HPN testbeds. We first identify a comprehensive list of attributes involved in a typical big data transfer process, including end host system configurations, network connection properties, and control parameters of data transfer methods. We then conduct an in-depth exploratory analysis of their impacts on application- level throughput, which provides insights into big data transfer performance and motivates the use of machine learning. We also investigate the applicability of machine learning algorithms and derive their general performance bounds for performance prediction of big data transfer in HPNs. Experimental results show that, with appropriate data preprocessing, the proposed machine learning-based approach achieves 95% or higher pre- diction accuracy in up to 90% of the cases with very noisy real-life performance measurements.

Yun, Daqing↗

Reliable extrapolation of deep neural operators informed by physics or sparse observations

Deep neural operators can learn nonlinear mappings between infinite-dimensional function spaces via deep neural networks. As promising surrogate solvers of partial differential equations (PDEs) for real-time prediction, deep neural operators such as deep operator networks (DeepONets) provide a new simulation paradigm in science and engineering. Pure data-driven neural operators and deep learning models, in general, are usually limited to interpolation scenarios, where new predictions utilize inputs within the support of the training set. However, in the inference stage of real-world applications, the input may lie outside the support, i.e., extrapolation is required, which may result to large errors and unavoidable failure of deep learning models. Here, we address this challenge of extrapolation for deep neural operators. First, we systematically investigate the extrapolation behavior of DeepONets by quantifying the extrapolation complexity, via the 2-Wasserstein distance between two function spaces and propose a new strategy of bias–variance trade-off for extrapolation with respect to model capacity. Subsequently, we develop a complete workflow, including extrapolation determination, and we propose five reliable learning methods that guarantee a safe prediction under extrapolation by requiring additional information—the governing PDEs of the system or sparse new observations. The proposed methods are based on either fine-tuning a pre-trained DeepONet or multifidelity learning. We demonstrate the effectiveness of the proposed framework for various types of parametric PDEs. Furthermore, our systematic comparisons provide practical guidelines for selecting a proper extrapolation method depending on the available information, desired accuracy, and required inference speed.

42 ENGINEERING↗

Structural constraint integration in a generative model for the discovery of quantum materials

Billions of organic molecules have been computationally generated, yet functional inorganic materials remain scarce due to limited data and structural complexity. Here, in this work, we introduce Structural Constraint Integration in a GENerative model (SCIGEN), a framework that enforces geometric constraints, such as honeycomb and kagome lattices, within diffusion-based generative models to discover stable quantum materials candidates. SCIGEN enables conditional sampling from the original distribution, preserving output validity while guiding structural motifs. This approach generates ten million inorganic compounds with Archimedean and Lieb lattices, over 10% of which pass multistage stability screening. High-throughput density functional theory calculations on 26,000 candidates shows over 95% convergence and 53% structural stability. A graph neural network classifier detects magnetic ordering in 41% of relaxed structures. Furthermore, we synthesize and characterize two predicted materials, TiPd 0.22 Bi 0.88 and Ti 0.5 Pd 1.5 Sb, which display paramagnetic and diamagnetic behaviour, respectively. Our results indicate that SCIGEN provides a scalable path for generating quantum materials guided by lattice geometry.

36 MATERIALS SCIENCE↗

Machine Learning for Daily Forecasts of Arctic Sea Ice Motion: An Attribution Assessment of Model Predictive Skill

Physics-based simulations of Arctic sea ice are highly complex, involving transport between different phases, length scales, and time scales. Resultantly, numerical simulations of sea ice dynamics have a high computational cost and model uncertainty. We employ data-driven machine learning (ML) to make predictions of sea ice motion. The ML models are built to predict present-day sea ice velocity given present-day wind velocity and previous-day sea ice concentration and velocity. Models are trained using reanalysis winds and satellite-derived sea ice properties. We compare the predictions of three different models: persistence (PS), linear regression (LR), and a convolutional neural network (CNN). We quantify the spatiotemporal variability of the correlation between observations and the statistical model predictions. Additionally, we analyze model performance in comparison to variability in properties related to ice motion (wind velocity, ice velocity, ice concentration, distance from coast, bathymetric depth) to understand the processes related to decreases in model performance. Results indicate that a CNN makes skillful predictions of daily sea ice velocity with a correlation up to 0.81 between predicted and observed sea ice velocity, while the LR and PS implementations exhibit correlations of 0.78 and 0.69, respectively. The correlation varies spatially and seasonally: lower values occur in shallow coastal regions and during times of minimum sea ice extent. LR parameter analysis indicates that wind velocity plays the largest role in predicting sea ice velocity on 1-day time scales, particularly in the central Arctic. Regions where wind velocity has the largest LR parameter are regions where the CNN has higher predictive skill than the LR.

54 ENVIRONMENTAL SCIENCES↗

Lambda-PFLOTRAN 1.0: a workflow for incorporating organic matter chemistry informed by ultra high resolution mass spectrometry into biogeochemical modeling

Abstract. Organic matter (OM) composition plays a central role in microbial respiration of dissolved organic matter and subsequent biogeochemical reactions. Here, a direct connection of organic matter chemistry and thermodynamics to reactive transport simulators has been achieved through the newly developed Lambda-PFLOTRAN workflow tool that succinctly incorporates carbon chemistry data generated from Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS) into reaction networks to simulate organic matter degradation and the resulting biogeochemistry. Lambda-PFLOTRAN is a Python-based workflow, executed through a Jupyter notebook interface, that digests raw FTICR-MS data, develops a representative reaction network based on substrate-explicit thermodynamic modeling (also termed lambda modeling due to its key thermodynamic parameter λ used therein), and completes a biogeochemical simulation with the open source, reactive flow and transport code PFLOTRAN. The workflow consists of the following five steps: configuration, thermodynamic (lambda) analysis, sensitivity analysis, parameter estimation, and simulation output and visualization. Two test cases are provided to demonstrate the functionality of the Lambda-PFLOTRAN workflow. The first test case uses laboratory incubation data of temporal oxygen depletion to fit lambda parameters (i.e., maximum utilization rate and microbial carrying capacity). A slightly more complex second test case fits multiple lambda formulation and soil organic matter release parameters to temporal greenhouse gas generation measured during a soil incubation. Overall, the Lambda-PFLOTRAN workflow facilitates upscaling by using molecular-scale characterization to inform biogeochemical processes occurring at larger scales.

58 GEOSCIENCES↗

Unleashing the Power of Wide-Area Data Analytics for Enhanced Grid Visibility

We report that the operation of modern electric transmission systems relies increasingly on the use of advanced sensing and synchronized-measurement technologies. These technologies collect and collate a wide variety of data from disperse locations to monitor the health of network, obtain fast and accurate diagnosis of systemic changes, and detect signs of anomalies and instabilities at an early stage. In view of the far-reaching blackout risk from potential damages to critical transmission grid assets (due to severe disturbances, physical attacks, and extreme weather events), stealthy activities that compromise the operation of wide-area sensing, communication, and control systems could cripple the entire cyber-physical smart-grid infrastructure. To this end, achieving robust event identification, ensuring bulk system observability, and developing actionable intelligence (by fostering data-driven decision making under imminent threat of complex, multifaceted attacks and disruptions) remain key challenges for sustaining a resilient power grid operation.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Architecture for Web-Based Visualization of Large-Scale Energy Domains: Preprint

With the growing penetration of inverter-based distributed energy resources and increased loads through electrification, power systems analyses are becoming more important and more complex. Moreover, these analyses increasingly involve the combination of interconnected energy domains with data that are spatially and temporally increasing in scale by orders of magnitude, surpassing the capabilities of many existing analysis and decision-support systems. We present the architectural design, development, and application of a high-resolution web-based visualization environment capable of cross-domain analysis of tens of millions of energy assets, focusing on scalability and performance. Our system supports the exploration, navigation, and analysis of large data from diverse domains such as electrical transmission and distribution systems, mobility and electric vehicle charging networks, communications networks, cyber assets, and other supporting infrastructure. We evaluate this system across multiple use cases, describing the capabilities and limitations of a web-based approach for high-resolution energy system visualizations.

grid modernization↗

A lightweight network anomaly detection technique

While the network anomaly detection is essential in network operations and management, it becomes further challenging to perform the first line of detection against the exponentially increasing volume of network traffic. In this paper, we develop a technique for the first line of online anomaly detection with two important considerations: (i) availability of traffic attributes during the monitoring time, and (ii) computational scalability for streaming data. The presented learning technique is lightweight and highly scalable with the beauty of approximation based on the grid partitioning of the given dimensional space. With the public traffic traces of KDD Cup 1999 and NSL-KDD, we show that our technique yields 98.5% and 83% of detection accuracy, respectively, only with a couple of readily available traffic attributes that can be obtained without the help of post-processing. Finally, the results are at least comparable with the classical learning methods including decision tree and random forest, with approximately two orders of magnitude faster learning performance.

97 MATHEMATICS AND COMPUTING↗

Artificial neural networks estimate evapotranspiration for Miscanthus × giganteus as effectively as empirical model but with fewer inputs

Estimating actual evapotranspiration (ET) is particularly crucial for addressing how vegetation affects the water balance of ecosystems. ET estimation can be complex with empirical models due to their many parameters and reliance on aridity. In contrast, artificial neural networks (ANNs) could potentially estimate ET with fewer and more common meteorological parameters. In this study, we trained two ANNs, one using a feed-forward approach (FFN) and the other a nonlinear auto-regressive network (NARX), to predict ET and compared them to the commonly used empirical model Granger and Gray (GG). We trained our models on a nine-year eddy covariance (EC) dataset for Miscanthu s × giganteus ( M . × giganteus ) from Illinois (UIEF), then tested them using out-of-sample data from both UIEF and a different location in Iowa (SABR) to compare the accuracy of FFN, NARX, and GG models in estimating daily ET. A combination of air temperature (T a ) and solar radiation (R s ) was chosen as inputs due to the highest R 2 for FFN (R 2 = 0.79, 0.81, and 0.79 for training, testing, and validation, respectively) and only T a for NARX (R 2 = 0.70 for out-of-sample validation). The predictive power of the FFN model was superior to the NARX and GG models at the UIEF site (R 2 = 0.84, 0.70, and 0.83 for out-of-sample validation, respectively). Our analysis showed that ANN approaches are as accurate as empirical approaches for estimating ET but use fewer inputs.

54 ENVIRONMENTAL SCIENCES↗

REVISITING THE SOLAR TACHOCLINE: AVERAGE PROPERTIES AND TEMPORAL VARIATIONS

The tachocline is believed to be the region where the solar dynamo operates. With over a solar cycle's worth of data available from the Michelson Doppler Imager and Global Oscillation Network Group instruments, we are in a position to investigate not merely the average structure of the solar tachocline, but also its time variations. We determine the properties of the tachocline as a function of time by fitting a two-dimensional model that takes latitudinal variations of the tachocline properties into account. We confirm that if we consider the central position of the tachocline, it is prolate. Our results show that the tachocline is thicker at latitudes higher than the equator, making the overall shape of the tachocline more complex. Of the tachocline properties examined, the transition of the rotation rate across the tachocline, and to some extent the position of the tachocline, show some temporal variations.

79 ASTRONOMY AND ASTROPHYSICS↗

Double Q-Learning for Radiation Source Detection

Anomalous radiation source detection in urban environments is challenging due to the complex nature of background radiation. When a suspicious area is determined, a radiation survey is usually carried out to search for anomalous radiation sources. To locate the source with high accuracy and in a short time, different survey approaches have been studied such as scanning the area with fixed survey paths and data-driven approaches that update the survey path on the fly with newly acquired measurements. In this work, we propose reinforcement learning as a data-driven approach to conduct radiation detection tasks with no human intervention. A simulated radiation environment is constructed, and a convolutional neural network-based double Q-learning algorithm is built and tested for radiation source detection tasks. Simulation results show that the double Q-learning algorithm can reliably navigate the detector and reduce the searching time by at least 44% compared with traditional uniform search methods and gradient search methods.

36 MATERIALS SCIENCE↗

sRecovery Act: Geologic Characterization of the South Georgia Rift Basin for Source Proximal CO 2 Storage

This study focuses on evaluating the feasibility and suitability of using the Jurassic/Triassic (J/TR) sediments of the South Georgia Rift basin (SGR) for CO 2 storage in southern South Carolina and southern Georgia The SGR basin in South Carolina (SC), prior to this project, was one of the least understood rift basin along the east coast of the U.S. In the SC part of the basin there was only one well (Norris Lightsey #1) the penetrated into J/TR. Because of the scarcity of data, a scaled approach used to evaluate the feasibility of storing CO 2 in the SGR basin. In the SGR basin, 240 km (~149 mi) of 2-D seismic and 2.6 km2 3-D (1 mi2) seismic data was collected, process, and interpreted in SC. In southern Georgia 81.3 km (~50.5 mi) consisting of two 2-D seismic lines were acquired, process, and interpreted. Seismic analysis revealed that the SGR basin in SC has had a very complex structural history resulting the J/TR section being highly faulted. The seismic data is southern Georgia suggest SGR basin has not gone through a complex structural history as the study area in SC. The project drilled one characterization borehole (Rizer # 1) in SC. The Rizer #1 was drilled but due to geologic problems, the project team was only able to drill to 1890 meters (6200 feet) instead of the proposed final depth 2744 meters (9002 feet). The drilling goals outlined in the original scope of work were not met. The project was only able to obtain 18 meters (59 feet) of conventional core and 106 rotary sidewall cores. All the conventional core and sidewall cores were in sandstone. We were unable to core any potential igneous caprock. Petrographic analysis of the conventional core and sidewall cores determined that the average porosity of the sedimentary material was 3.4% and the average permeability was 0.065 millidarcy. Compaction and diagenetic studies of the samples determined there would not be any porosity or permeability at depth in SC. In Georgia there appears to be porosity in the J/TR section based on neutron log porosity values. The only zones in Rizer #1 that appear to be porous were fractured diabase units where saline formation water was flowing into the borehole. Two geocellular models were created for the SC and GA study area. Flow simulation modeling was performed on the SC data set. The injection simulation used the newly acquired basin data as well as the Petrel 3-D geologic model that included geologic structure. Due to the new basin findings as a result of the newly acquired data, during phase two of the modeling the diabase unit was used as reservoir and the sandstone units were used as caprock. Conclusion are: 1) the SGR basin is composed of numerous sub-basins, 2) this study only looked at portions of two sub-basins, 3) in SC, 30 million tonnes of CO 2 can be injected into the diabase units if the fracture network is continuous through the units, 4) due to the severity of the faulting there is no way of assuring the injected CO 2 will not migrate upward into the overlying Coastal Plain aquifers, 5) in Georgia there appears to porous zones in the J/TR sandstones, 6) as in SC there is faulting in the sub-basin and the seismic suggest the faulting extends upward into the Coastal Plain making that area not suitable for CO 2 sequestration, 7) the complex faulting observed at both study areas appear to be associated with transfer fault zones (Heffner 2013), if sub-basins in the Georgia portion of the SGR basin can be located that are far away from the transfer fault zones there is a strong possibility of sequestering CO 2 in these areas, and 9) the SGR basin covers area in three states and this project only studied two small areas so there is enormous potential for CO 2 sequestration in other portions the basin and further research needs to be done to find these areas.

58 GEOSCIENCES↗

Revealing the unusual rate-dependent mechanical behaviors of nematic liquid crystal elastomers

Liquid crystal elastomers (LCEs) exhibit unique mechanical properties of soft elasticity and enhanced energy dissipation with rate dependency. They are potentially transformative materials for applications in mechanical impact mitigation and vibration isolation. However, previous studies have primarily focused on the mechanics of LCEs under equilibrium and quasistatic loading conditions. Critical knowledge gaps exist in understanding their rate-dependent behaviors, which are a complex mixture of traditional network viscoelasticity and the soft elastic behaviors with changes in the mesogen orientation and order parameter. Together, these inelastic mechanisms lead to unusual rate-dependent energy absorption responses of LCEs. In this work, we developed a viscoelastic constitutive theory for monodomain nematic LCEs to investigate how multiple underlying sources of inelasticity manifest in the rate-dependent and dissipative behaviors of monodomain LCEs. Here, the theoretical modeling framework combines the neo-classical network theory with evolution rules for the mesogen orientation and order parameter with conventional viscoelasticity. The model is calibrated with uniaxial tension and compression data spanning six decades of strain rates. The established 3D constitutive model enables general loading predictions taking the initial mesogen orientation and order parameter as inputs. Additionally, parametric studies were performed to further understand the rate dependence of monodomain LCEs in relation to their energy absorption characteristics. Based on the parametric studies, particularly loading scenarios are identified as conditions where LCEs outperform conventional elastomers regarding energy absorption.

42 ENGINEERING↗

Explicit physics-informed neural networks for nonlinear closure: The case of transport in tissues

In upscaling methods, closures for nonlinear problems present a well-known challenge. While a number of theoretical methods have been proposed for handling such closures, nonlinearities still remain a significant obstacle for many problems. In this work, we use a combination of formal upscaling and data-driven machine learning for explicitly closing a nonlinear transport and reaction process in multiscale tissues. The classical effectiveness factor model is used to formulate the macroscale reaction kinetics. We train a multilayer perceptron network using training data generated by direct numerical simulations over microscale examples. Once trained, the network is used in an algorithm for numerically solving the upscaled (coarse-grained) differential equation describing mass transport and reaction in two example tissues. The network is described as being explicit in the sense that the network is trained using macroscale concentrations and gradients of concentration as components of the feature space rather than incorporating them as part of a constraint in the optimization process. Network training and solutions to the macroscale transport equations were computed for two different tissues. The two tissue types (brain and liver) exhibit markedly different geometrical complexity and spatial scale (cell size and sample size). The upscaled solutions for the average concentration are compared with numerical solutions derived from the microscale concentration fields by a posteriori averaging. There are three outcomes of this work of particular note. 1) Our overall approach results in an upscaled nonlinear PDE. The PDE is closed using a neural network, and our approach results in the definition of the classical effectiveness factor for effecting closure. 2) We identify particular source terms for the closure problem that are important for representing the structure of the closure. These source terms involve macroscale concentrations and their gradients. We adopt these source terms to use as explicit features in the learning algorithm. We find the trained networks that include the macroscale source terms generate models that are able to predict the correction factor with increased fidelity over those that do not. 3) We find that the trained network exhibits good generalizability, and it is able to predict the effectiveness factor with high fidelity for realistically-structured tissues despite the significantly different scale and geometrical complexity of the two example tissue types. This latter result emphasizes our purposeful connection between conventional averaging methods with the use of machine learning for closure; this contrasts with some machine learning methods for upscaling where the exact form of the macroscale equation remains unknown.

97 MATHEMATICS AND COMPUTING↗

Deep Learning for Semantic Segmentation of Defects in Advanced STEM Images of Steels

Crystalline materials exhibit long-range ordered lattice unit, within which resides nonperiodic structural features called defects. These crystallographic defects play a vital role in determining the physical and mechanical properties of a wide range of material systems. While computer vision has demonstrated success in recognizing feature patterns in images with well defined contrast, automated identification of nanometer scale crystallographic defects in electron micrographs governed by complex contrast mechanism is still a challenging task. Here, building upon an advanced defect imaging mode that offers high feature clarity, we introduce DefectSegNet - a new convolutional neural network (CNN) architecture that performs semantic segmentation of three common crystallographic defects in structural alloys: dislocation lines, precipitates and voids. Results from supervised training of a small set of high-quality defect images of steels show high pixel-wise accuracy across all three types of defects: 91.60 ± 1.77% on dislocations, 93.39 ± 1.00% on precipitates, and 98.85 ± 0.56% on voids. We discuss the sources of uncertainties in CNN prediction, and the training data in terms of feature density, representation and homogeneity and their effects on deep learning performance. Further defect quantification using DefectSegNet prediction outperforms human expert average, presenting a promising new workflow for fast and statistically meaningful quantification of materials defects.

36 MATERIALS SCIENCE↗

Addressing Data Veracity in Big Data Applications

Big data applications such as in smart electric grids, transportation, and remote environment monitoring involve geographically dispersed sensors that periodically send back information to central nodes. In many cases, data from sensors is not available at central nodes at a frequency that is required for real-time modeling and decision-making. This may be due to physical limitations of the transmission networks, or due to consumers limiting frequent transmission of data from sensors located at their premises for security and privacy concerns. Such scenarios lead to partial data problem and raise the issue of data veracity in big data applications. We describe a novel solution to the problem of making short term predictions (up to a few hours ahead) in absence of real-time data from sensors in Smart Grid. A key implication of our work is that by using real-time data from only a small subset of influential sensors, we are able to make predictions for all sensors. We thus reduce the communication complexity involved in transmitting sensory data in Smart Grids. We use real-world electricity consumption data from smart meters to empirically demonstrate the usefulness of our method. Our dataset consists of data collected at 15-min intervals from 170 smart meters in the USC Microgrid for 7 years, totaling 41,697,600 data points.

97 MATHEMATICS AND COMPUTING↗

Ps and Qs: Quantization-Aware Pruning for Efficient Low Latency Neural Network Inference

Efficient machine learning implementations optimized for inference in hardware have wide-ranging benefits, depending on the application, from lower inference latency to higher data throughput and reduced energy consumption. Two popular techniques for reducing computation in neural networks are pruning, removing insignificant synapses, and quantization, reducing the precision of the calculations. In this work, we explore the interplay between pruning and quantization during the training of neural networks for ultra low latency applications targeting high energy physics use cases. Techniques developed for this study have potential applications across many other domains. We study various configurations of pruning during quantization-aware training, which we term quantization-aware pruning, and the effect of techniques like regularization, batch normalization, and different pruning schemes on performance, computational complexity, and information content metrics. We find that quantization-aware pruning yields more computationally efficient models than either pruning or quantization alone for our task. Further, quantization-aware pruning typically performs similar to or better in terms of computational efficiency compared to other neural architecture search techniques like Bayesian optimization. Surprisingly, while networks with different training configurations can have similar performance for the benchmark application, the information content in the network can vary significantly, affecting its generalizability.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗