Search NASA⌕ Search

SEARCH · Search NASA

Results for “large scale 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.

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

At least 469 records · Page 26
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Experiments to Quantify the Shielding Efficacy of Materials to be Implemented for Portable Shielding Enclosures that Adequately Mitigate EMI from Lighting Attachment Events

Mitigating electric and magnetic fields induced by lightning strike events is critical for the security and longevity of electrical components. Electromagnetic interference originating from a lightning strike is categorized into the low to medium frequency ranges, commonly referred to as the “magnetic coupling” regime. This low-frequency characteristic brings about the problem of shielding magnetic flux lines. Such shielding can be accomplished with highly conductive materials via the generation of eddy currents or by diverting the flux lines using materials with a large relative permeability. The shielding effectiveness of materials with varying thicknesses and geometry are tested using a pulse forming network having comparable characteristics to a scaled lighting strike. A four-stage Marx generator with a peak current of ~33 kA and a risetime of 1.4 μs is used as the pulse forming network to emulate the scaled electric and magnetic fields. The data collected from these experiments provide information and techniques that can be implemented into the design of a portable, deployable shielding enclosure.

61 RADIATION PROTECTION AND DOSIMETRY↗
DOE OSTI2021

DETECTING FIRE WITH MACHINE LEARNING-ENABLED VISUAL MONITORING FOR NUCLEAR POWER PLANT ENVIRONMENTS

Nuclear power plants are experiencing significant cost challenges to remain competitive with other energy-generation utilities. Unlike other industries, the cost of operation and maintenance activities is mostly attributed to workforce costs. To mitigate this, nuclear power plant stakeholders are increasingly interested in the development and deployment of machine learning methods to potentially automate or augment manually intensive tasks to reduce costs, especially for monitoring activities. One monitoring function that is visually demanding and that can occur frequently to meet the requirements of a fire protection program is visually monitoring an area for fire occurrence. Currently, fire watch activities consist of a worker physically stationed at a given location with the sole responsibility of observing a given area to ensure a fire is detected and mitigated promptly. This effort focused on the development and evaluation of a suitable deep convolutional neural network to classify individual video frames at a sub-second frequency for the occurrence of “fire” and “no fire” in varying industrial environments similar to nuclear power plants. It is believed that a trained neural network model could be integrated with existing facility video surveillance camera feeds to generate alerts when fire inferences occur in individual frames captured at sub-second temporal resolutions. Extensive effort was dedicated to identifying and curating suitable imagery training data representing varying environments and scene settings with and without flame features to maximize generalization in nuclear power plant environments. The data collection effort resulted in the aggregation of a large, labeled image library exceeding 12,000 images to support model training for diverse industrial environments. A deep neural network model incorporating parallel multi-scale capabilities was developed and trained to support accurate image-based detection of flame incidents of varying sizes and spectral feature properties within heterogeneous scenes. Analysis results show that the trained model can achieve high inference accuracy despite heterogeneous scene environments and components. Testing accuracy exceeded 95.0 percent with very low false positive and false negative inferences.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

CO 2 Transport Infrastructure Outlook in the United States

Carbon capture and storage (CCS) represents one of the most important methods to mitigate anthropogenic carbon emissions at a large scale, playing a key role in meeting climate change targets (Bui et al., 2018) and for net-zero CO 2 by 2050 scenarios in the United States (Browning et al., 2023). This technology involves capturing CO 2 emissions from industrial processes, transporting them via pipelines, trucks, rails, or ships, and ultimately storing them in underground geological sites, such as saline aquifers or depleted oil reservoirs. Thus, to encourage carbon reduction initiatives, the U.S. Congress enacted the Bipartisan Budget Act in 2018, reforming the 45Q tax credit to benefit operators storing CO 2 in geologic formations (Jones and Sherlock, 2021). Additionally, the 2022 Inflation Reduction Act further expanded these incentives, providing additional support for CCS initiatives (Hackett and Kuehn, 2023). Although numerous studies describe the importance of optimal CO 2 transportation to support the decision-making of CCS projects aligned with the objective of net-zero emissions by 2050 (Abramson and Christensen, 2021; Chen and Pawar, 2023; Greig and Pascale, 2021), further efforts are required to optimize the transport infrastructure for national-scale CCS deployment. Therefore, in this study, we examine three nationwide scenarios with the SimCCS 3.0 tool (Ma et al., 2022, 2023, 2024) along with a novel geospatial splitting approach developed by Velasco-Lozano et al. (Velasco- Lozano et al., 2024a, 2024b). We present optimized pipeline networks that meet the dynamic evolution of annual capture amounts, describing the required total pipeline lengths at each stage as a function of the pipeline diameters. Thus, the cases presented demonstrate the feasibility of CO 2 pipeline infrastructure for large-scale CCS projects.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Distributed deep learning training using silicon photonic switched architectures

The scaling trends of deep learning models and distributed training workloads are challenging network capacities in today’s datacenters and high-performance computing (HPC) systems. We propose a system architecture that leverages silicon photonic (SiP) switch-enabled server regrouping using bandwidth steering to tackle the challenges and accelerate distributed deep learning training. In addition, our proposed system architecture utilizes a highly integrated operating system-based SiP switch control scheme to reduce implementation complexity. To demonstrate the feasibility of our proposal, we built an experimental testbed with a SiP switch-enabled reconfigurable fat tree topology and evaluated the network performance of distributed ring all-reduce and parameter server workloads. The experimental results show up to 3.6× improvements over the static non-reconfigurable fat tree. Our large-scale simulation results show that server regrouping can deliver up to 2.3× flow throughput improvement for a 2× tapered fat tree and a further 11% improvement when higher-layer bandwidth steering is employed. The collective results show the potential of integrating SiP switches into datacenters and HPC systems to accelerate distributed deep learning training.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2025

Leveraging operator learning to accelerate convergence of the preconditioned conjugate gradient method

We propose a new deflation strategy to accelerate the convergence of the preconditioned conjugate gradient (PCG) method for solving parametric large-scale linear systems of equations. Unlike traditional deflation techniques that rely on eigenvector approximations or recycled Krylov subspaces, we generate the deflation subspaces using operator learning, specifically the Deep Operator Network (DeepONet). To this aim, we introduce two complementary approaches for assembling the deflation operators. The first approach approximates near-null space vectors of the discrete PDE operator using the basis functions learned by the DeepONet. The second approach directly leverages solutions predicted by the DeepONet. To further enhance convergence, we also propose several strategies for prescribing the sparsity pattern of the deflation operator. Here, a comprehensive set of numerical experiments encompassing steady-state, time-dependent, scalar, and vector-valued problems posed on both structured and unstructured geometries is presented and demonstrates the effectiveness of the proposed DeepONet-based deflated PCG method, as well as its generalization across a wide range of model parameters and problem resolutions.

Deflation↗
DOE OSTI2025

Enhancing 2D hydrodynamic flood models through machine learning and urban drainage integration

Two-dimensional hydrodynamic flood models are commonly employed for simulating flood extent and inundation depth. However, the influence of urban drainage network (UDN) is frequently overlooked in these models, potentially compromising their accuracy. Furthermore, the expensive computational costs and longer processing times make them challenging for large-scale hydrodynamic simulation. To address these challenges, this paper develops a machine learning (ML)-driven emulator for an open-source flood model, the Two-dimensional Runoff Inundation Toolkit for Operational Needs (TRITON). A TRITON-ML Emulator (TR-Emulator) that utilizes Convolutional Long Short-Term Memory is developed to capture the spatiotemporal features of flood events based on the outputs from TRITON. We further enhance the emulator by integrating UDN parameters (TR-UDN), such as the flow capacity of drainage pipes, pipe size, and pipe length, via an ML stacking technique to improve the water surface elevation (WSE) simulation. Hurricane Harvey 2017 in Houston, TX is used as the case study. We compare WSE results from TRITON, TR-Emulator, TR-UDN, and the United States Geological Survey (USGS) observations to evaluate the performance of these models. The results indicate that the TR-Emulator effectively replicates the WSE simulated by TRITON. Additionally, TR-UDN performs well in capturing WSE patterns and peak flows, aligning more closely with USGS observations, except in areas with milder slopes where conveyance discrepancies are observed. We further test the generalizability of our ML-based models using another smaller event. This paper shows that the TR-Emulator is effective for users and engineers to emulate a 2D hydrodynamic model, and the enhanced version of the TR-Emulator, TR-UDN, can be an efficient tool for predicting WSEs during urban flooding.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Tracking seasonal variability in plant traits from spaceborne PRISMA and NEON AOP across forest types and ecoregions

Plant traits serve as critical indicators of how plants adapt to environmental changes and influence ecosystem functions. While airborne hyperspectral remote sensing effectively maps plant traits through detailed reflectance properties, it is limited by cost and scale, making large-scale and temporal studies challenging. The recently launched spaceborne hyperspectral imager, PRecursore IperSpettrale della Missione Applicativa (PRISMA), offers frequent, large scale and high-fidelity observations on a spatial resolution of 30 m and a revisit time of around 29 days, making it suitable for large-scale seasonal trait mapping. However, their potential remains largely unexplored. This study developed a multi-stage framework by leveraging the PRISMA spaceborne hyperspectral data and National Ecological Observatory Network (NEON) Airborne Observation Platform (AOP) hyperspectral data to investigate the seasonal dynamics of four key plant traits — chlorophyll content, carotenoid content, equivalent water thickness, and nitrogen content — across eleven NEON sites representing diverse forest types and ecoregions in the contiguous U.S. Our results demonstrated that PRISMA hyperspectral data can reliably track seasonal variability in plant traits, achieving overall R 2 values ranging from 0.78 to 0.88 and normalized root mean square error (NRMSE) values ranging from 5.4% to 8.4% for the four traits. Seasonal patterns revealed bell-shaped trajectories for chlorophyll and carotenoids, while equivalent water thickness decreased steadily across most sites, driven by structural changes during leaf maturation and senescence. Nitrogen content exhibited less pronounced seasonal variation but followed expected nutrient resorption patterns. Analysis of environmental drivers showed that seasonal variability is primarily controlled by solar radiation and day length in northern sites, vapor pressure in semi-arid regions, and temperature in mid-southeastern sites. Spatial variability, meanwhile, was primarily driven by soil properties, particularly during the peak growing season. However, the influence of soil variables slightly declines toward the end of the season at several sites, as climatic factors become more prominent. This study highlights the capability of PRISMA, and potentially other similar spaceborne hyperspectral data for large-scale, time-series plant trait mapping and provides valuable insights into the interactions between plant traits and environmental factors. In conclusion, these findings contribute to advancing our understanding of plant functional ecology and improving predictions of ecosystem responses to environmental changes.

Environmental drivers↗
DOE OSTI2020

Neural networks for modeling and control of particle accelerators

Charged particle accelerators support a wide variety of scientific, industrial, and medical applications. They range in scale and complexity from systems with just a few components for beam acceleration and manipulation, to large scientific user facilities that span many kilometers and have hundreds-to-thousands of individually-controllable components. Specific operational requirements must be met by adjusting the many controllable variables of the accelerator. Meeting these requirements can be challenging, both in terms of the ability to achieve specific beam quality metrics in a reliable fashion and in terms of the time needed to set up and maintain the optimal operating conditions. One avenue toward addressing this challenge is to incorporate techniques from the fields of machine learning (ML) and artificial intelligence (AI) into the way particle accelerators are modeled and controlled. While many promising approaches within AI/ML could be used for particle accelerators, this dissertation focuses on approaches based on neural networks. Neural networks are particularly well-suited to modeling, control, and diagnostic analysis of nonlinear systems, as well as systems with large parameter spaces. They are also very appealing for their ability to process high-dimensional data types, such as images and time series (both of which are ubiquitous in particle accelerators). In this work, key studies that demonstrated the potential utility of modern neural network-based approaches to modeling and control of particle accelerators are presented. The context for this work is important: at the start of this work in 2012, there was little interest in AI/ML in the particle accelerator community, and many of the advances in neural networks and deep learning that enabled its present success had not yet been made at that time. As such, this work was both an exploration of possible application areas and a generator of proof-of-concept demonstrations in these areas.

43 PARTICLE ACCELERATORS↗
DOE OSTI2025

The Statistical Spread of Transmission Outages on a Fast Protection Time Scale Based on Utility Data

When there is a fault, the protection system automatically removes one or more transmission lines on a fast time scale of less than one minute. The outaged lines form a pattern in the transmission network. We extract these patterns from utility outage data, determine some key statistics of these patterns, and then show how to generate new patterns consistent with these statistics. The generated patterns provide a new and easily feasible way to model the overall effect of the protection system at the scale of a large transmission system. This new data-driven generative modeling of protection is expected to contribute to simulations of disturbances in large grids so that they can better quantify the risk of blackouts. Analysis of the pattern sizes suggests an index that describes how much outages spread in the transmission network at the fast timescale.

Transmission↗
DOE OSTI2022

Scalable deep learning for watershed model calibration

Watershed models such as the Soil and Water Assessment Tool (SWAT) consist of high-dimensional physical and empirical parameters. These parameters often need to be estimated/calibrated through inverse modeling to produce reliable predictions on hydrological fluxes and states. Existing parameter estimation methods can be time consuming, inefficient, and computationally expensive for high-dimensional problems. In this paper, we present an accurate and robust method to calibrate the SWAT model (i.e., 20 parameters) using scalable deep learning (DL). We developed inverse models based on convolutional neural networks (CNN) to assimilate observed streamflow data and estimate the SWAT model parameters. Scalable hyperparameter tuning is performed using high-performance computing resources to identify the top 50 optimal neural network architectures. We used ensemble SWAT simulations to train, validate, and test the CNN models. We estimated the parameters of the SWAT model using observed streamflow data and assessed the impact of measurement errors on SWAT model calibration. We tested and validated the proposed scalable DL methodology on the American River Watershed, located in the Pacific Northwest-based Yakima River basin. Our results show that the CNN-based calibration is better than two popular parameter estimation methods (i.e., the generalized likelihood uncertainty estimation [GLUE] and the dynamically dimensioned search [DDS], which is a global optimization algorithm). For the set of parameters that are sensitive to the observations, our proposed method yields narrower ranges than the GLUE method but broader ranges than values produced using the DDS method within the sampling range even under high relative observational errors. The SWAT model calibration performance using the CNNs, GLUE, and DDS methods are compared using R 2 and a set of efficiency metrics, including Nash-Sutcliffe, logarithmic Nash-Sutcliffe, Kling-Gupta, modified Kling-Gupta, and non-parametric Kling-Gupta scores, computed on the observed and simulated watershed responses. The best CNN-based calibrated set has scores of 0.71, 0.75, 0.85, 0.85, 0.86, and 0.91. The best DDS-based calibrated set has scores of 0.62, 0.69, 0.8, 0.77, 0.79, and 0.82. The best GLUE-based calibrated set has scores of 0.56, 0.58, 0.71, 0.7, 0.71, and 0.8. The scores above show that the CNN-based calibration leads to more accurate low and high streamflow predictions than the GLUE and DDS sets. Our research demonstrates that the proposed method has high potential to improve our current practice in calibrating large-scale integrated hydrologic models.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

Transferring predictions of formation energy across lattices of increasing size*

In this study, we show the transferability of graph convolutional neural network (GCNN) predictions of the formation energy of the nickel-platinum solid solution alloy across atomic structures of increasing sizes. The original dataset was generated with the large-scale atomic/molecular massively parallel simulator using the second nearest-neighbor modified embedded-atom method empirical interatomic potential. Geometry optimization was performed on the initially randomly generated face centered cubic crystal structures and the formation energy has been calculated at each step of the geometry optimization, with configurations spanning the whole compositional range. Using data from various steps of the geometry optimization, we first trained our open-source, scalable implementation of GCNN called HydraGNN on a lattice of 256 atoms, which accounts well for the short-range interactions. Using this data, we predicted the formation energy for lattices of 864 atoms and 2048 atoms, which resulted in lower-than-expected accuracy due to the long-range interactions present in these larger lattices. We accounted for the long-range interactions by including a small amount of training data representative for those two larger sizes, whereupon the predictions of HydraGNN scaled linearly with the size of the lattice. Therefore, our strategy ensured scalability while reducing significantly the computational cost of training on larger lattice sizes.

36 MATERIALS SCIENCE↗
DOE OSTI2022

A Comparison of Linear Solvers for Resolving Flow in Three-Dimensional Discrete Fracture Networks

We compare various methods for resolving steady flow within three-dimensional discrete fracture networks, including direct methods, Krylov subspace methods with and without preconditioning, and multi-grid methods. We compared the performance of the methods based on compute times and scaling of the solution as a function of the number of grid nodes and log-variance of the hydraulic aperture. The methods are applied to three test cases: (a) variable density of networks with a truncated power-law distribution of fracture lengths, (b) a fixed network composed of monodisperse fracture sizes but varied permeability/aperture heterogeneity, (c) and a network based on field site in Nevada, US. We chose these cases to allow us to study the impact of the mesh size and flow properties, as well as to demonstrate our conclusions on a large-scale, realistic problem (more than 40 million mesh nodes). A direct solution using Cholesky factorization outperformed other methods for every example but was closely followed in performance by some algebraic multigrid (AMG) preconditioned Krylov subspace methods. Among the Krylov methods, conjugate gradients (CG) with an AMG preconditioner performs the best. Generally, Cholesky factorization is recommended, but CG with an AMG preconditioner may be suitable for very large problems beyond 40 million nodes where the entire linear system cannot reside in memory.

58 GEOSCIENCES↗
DOE OSTI2024

MDLoader: A Hybrid Model-Driven Data Loader for Distributed Graph Neural Network Training

Scalable data management is essential for processing large scientific dataset on HPC platforms for distributed deep learning. In-memory distributed storage is preferred for its speed, enabling rapid, random, and frequent data access required by stochastic optimizers. Processes use one-sided or collective communication to fetch remote data, with optimal performance depending on (i) dataset characteristics, (ii) training scale, and (iii) interconnection network. Empirical analysis shows collective communication excels with larger mini-batch sizes and/or fewer processes, whereas one-sided communication outperforms at larger scales. We propose MDLoader, a hybrid in-memory data loader for distributed graph neural network training. MDLoader features a model-driven performance estimator that dynamically selects between one-sided and collective communication at the beginning of training using Tree of Parzen Estimators (TPE). Evaluations on NERSC Perlmutter and OLCF Summit show MDLoader outperforms single-backend loaders by up to 2.83 × and predicts the suitable communication method with 96.3% (Perlmutter) and 94.3% (Summit) success rate.

Bae, Jonghyun↗
DOE OSTI2020

Learning to Count Grave Sites for Cemetery Observation Models With Satellite Imagery

Understanding how people occupy open spaces is important for research in support of population modeling, policy, national security, emergency response, and sustainability. For the past decade, there has been an increase in research toward capturing and reporting population dynamics and patterns of life at the building level and in some open public spaces such as cemeteries and parks. This is done through observation models developed from local sociocultural information acquired at various spatiotemporal scales to inform night, day, and episodic population occupancy estimates (people/1000 sq ft). Sociocultural information for cemeteries and parks is scarcely available and often collected manually. The process is not only marred by inconsistencies but is laborious and time consuming. In this study, we leverage convolutional neural networks (CNNs) and satellite imagery to derive grave site counts as proxy variables to support scalable and accurate sociocultural data required in a population observation model. Through a hybrid workflow (weak localization plus regression model), we characterize a large scale automation process to counting of grave sites. We evaluate and demonstrate the efficacy of proposed workflow using out-of-data set large satellite imagery and establish its broader impact on cemetery observation models.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗
DOE OSTI2023

Architecture for Co-Simulation of Transportation and Distribution Systems with Electric Vehicle Charging at Scale in the San Francisco Bay Area

This work describes the Grid-Enhanced, Mobility-Integrated Network Infrastructures for Extreme Fast Charging (GEMINI) architecture for the co-simulation of distribution and transportation systems to evaluate EV charging impacts on electric distribution systems of a large metropolitan area and the surrounding rural regions with high fidelity. The current co-simulation is applied to Oakland and Alameda, California, and in future work will be extended to the full San Francisco Bay Area. It uses the HELICS co-simulation framework to enable parallel instances of vetted grid and transportation software programs to interact at every model timestep, allowing high-fidelity simulations at a large scale. This enables not only the impacts of electrified transportation systems across a larger interconnected collection of distribution feeders to be evaluated, but also the feedbacks between the two systems, such as through control systems, to be captured and compared. The findings are that with moderate passenger EV adoption rates, inverter controls combined with some distribution system hardware upgrades can maintain grid voltages within ANSI C.84 range A limits of 0.95 to 1.05 p.u. without smart charging. However, EV charging control may be required for higher levels of charging or to reduce grid upgrades, and this will be explored in future work.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2020

Large-scale proteomic analysis of Alzheimer’s disease brain and cerebrospinal fluid reveals early changes in energy metabolism associated with microglia and astrocyte activation

Our understanding of Alzheimer’s disease (AD) pathophysiology remains incomplete. Here, for this work, we used quantitative mass spectrometry and coexpression network analysis to conduct the largest proteomic study thus far on AD. A protein network module linked to sugar metabolism emerged as one of the modules most significantly associated with AD pathology and cognitive impairment. This module was enriched in AD genetic risk factors and in microglia and astrocyte protein markers associated with an anti-inflammatory state, suggesting that the biological functions it represents serve a protective role in AD. Proteins from this module were elevated in cerebrospinal fluid in early stages of the disease. In this study of >2,000 brains and nearly 400 cerebrospinal fluid samples by quantitative proteomics, we identify proteins and biological processes in AD brains that may serve as therapeutic targets and fluid biomarkers for the disease.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2025

Tensile Modeling PVC Gels for Electrohydraulic Actuators

Polyvinyl chloride (PVC)-dibutyl adipate (DBA) gels are a fascinating dielectric elastomer actuator showing promise in soft robotics. When actuated with high voltages, the gel deforms towards the anode. A recent application of PVC gels in electrohydraulic actuators motivates elastic and hyperelastic constitutive relationships for tensile loading modes. PVC gels with plasticizer-to-polymer weight ratios of 2:1, 4:1, 6:1, and 8:1 w/w were evaluated. PVC gels exhibit a linear elastic region up to 25% strain. The elastic modulus decreased with increasing plasticizer content from 288.8 kPa, 56.1 kPa, 24.7 kPa, to 11 kPa. Poisson’s ratio also decreased with increasing plasticizer content from 0.42, 0.43, 0.39, to 0.35. We suggest that the decrease in polymer concentration facilitates a weakly interconnected polymer network susceptible to chain slippage that hinders the network response, thus lowering Poisson’s ratio. Our work suggests that PVC gels can be treated as isotropic and incompressible for large strains and hyperelastic modeling; however, highly plasticized gels tend to act less incompressible at small strains. The power scaling law between the elastic modulus and plasticizer weight ratio showed high agreement, making the elastic modulus deterministic for any plasticizer content. The Neo–Hookean, Mooney–Rivlin, Yeoh, Gent, Ogden, and extended tube hyperelastic constitutive models are investigated. The Yeoh model shows the highest feasibility when evaluated up to 3.5 stretch, showing a maximum normalized root-mean-square-error of 6.85%. Together, these findings establish a constitutive basis for PVC-DBA gels, incorporating small strain elasticity, large strain non-linear behavior, and network analysis while providing suggestive insight into the network structure required for accurately modeling the EPIC.

Polymer Science↗
DOE OSTI2022

Trainability of Dissipative Perceptron-Based Quantum Neural Networks

Several architectures have been proposed for quantum neural networks (QNNs), with the goal of efficiently performing machine learning tasks on quantum data. Rigorous scaling results are urgently needed for specific QNN constructions to understand which, if any, will be trainable at a large scale. Here, we analyze the gradient scaling (and hence the trainability) for a recently proposed architecture that we call dissipative QNNs (DQNNs), where the input qubits of each layer are discarded at the layer’s output. We find that DQNNs can exhibit barren plateaus, i.e., gradients that vanish exponentially in the number of qubits. Moreover, we provide quantitative bounds on the scaling of the gradient for DQNNs under different conditions, such as different cost functions and circuit depths, and show that trainability is not always guaranteed. Here our work represents the first rigorous analysis of the scalability of a perceptron-based QNN.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗