Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “neural network architecture”

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 955 records · Page 53
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Graphene U-Net v1

Graphene U-Net is a library that provides a simplified platform for training deep neural networks for the task of microscopy image segmentation. It contains functions and classes that make the process of training a neural network simple such that non-ML experts can train and evaluate models on their own datasets. It uses Pytorch as a backend and can run on both CUDA-enable GPUs and CPUs. It contains the main library file Microscopy_Unet.py as well as unet.py which contains the UNET model used for this software. This can be replaced with any other fully convolutional deep learning architecture with relative ease.

Sadre, Robbie↗
DOE OSTI2023

Data augmentation and multimodal learning for predicting drug response in patient-derived xenografts from gene expressions and histology images

Patient-derived xenografts (PDXs) are an appealing platform for preclinical drug studies. A primary challenge in modeling drug response prediction (DRP) with PDXs and neural networks (NNs) is the limited number of drug response samples. We investigate multimodal neural network (MM-Net) and data augmentation for DRP in PDXs. The MM-Net learns to predict response using drug descriptors, gene expressions (GE), and histology whole-slide images (WSIs). We explore whether combining WSIs with GE improves predictions as compared with models that use GE alone. We propose two data augmentation methods which allow us training multimodal and unimodal NNs without changing architectures with a single larger dataset: 1) combine single-drug and drug-pair treatments by homogenizing drug representations, and 2) augment drug-pairs which doubles the sample size of all drug-pair samples. Unimodal NNs which use GE are compared to assess the contribution of data augmentation. The NN that uses the original and the augmented drug-pair treatments as well as single-drug treatments outperforms NNs that ignore either the augmented drug-pairs or the single-drug treatments. In assessing the multimodal learning based on the MCC metric, MM-Net outperforms all the baselines. Our results show that data augmentation and integration of histology images with GE can improve prediction performance of drug response in PDXs.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2024

Deep learning models for interpretation of point of care ultrasound in military working dogs

Introduction: Military working dogs (MWDs) are essential for military operations in a wide range of missions. With this pivotal role, MWDs can become casualties requiring specialized veterinary care that may not always be available far forward on the battlefield. Some injuries such as pneumothorax, hemothorax, or abdominal hemorrhage can be diagnosed using point of care ultrasound (POCUS) such as the Global FAST® exam. This presents a unique opportunity for artificial intelligence (AI) to aid in the interpretation of ultrasound images. In this article, deep learning classification neural networks were developed for POCUS assessment in MWDs. Methods: Images were collected in five MWDs under general anesthesia or deep sedation for all scan points in the Global FAST® exam. For representative injuries, a cadaver model was used from which positive and negative injury images were captured. A total of 327 ultrasound clips were captured and split across scan points for training three different AI network architectures: MobileNetV2, DarkNet-19, and ShrapML. Gradient class activation mapping (GradCAM) overlays were generated for representative images to better explain AI predictions. Results: Performance of AI models reached over 82% accuracy for all scan points. The model with the highest performance was trained with the MobileNetV2 network for the cystocolic scan point achieving 99.8% accuracy. Across all trained networks the diaphragmatic hepatorenal scan point had the best overall performance. However, GradCAM overlays showed that the models with highest accuracy, like MobileNetV2, were not always identifying relevant features. Conversely, the GradCAM heatmaps for ShrapML show general agreement with regions most indicative of fluid accumulation. Discussion: Overall, the AI models developed can automate POCUS predictions in MWDs. Preliminarily, ShrapML had the strongest performance and prediction rate paired with accurately tracking fluid accumulation sites, making it the most suitable option for eventual real-time deployment with ultrasound systems. Further integration of this technology with imaging technologies will expand use of POCUS-based triage of MWDs.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

Strong Lens Discoveries in DESI Legacy Imaging Surveys DR10 with Two Deep Learning Architectures

Abstract We have conducted a search for strong gravitational lensing systems in the Dark Energy Spectroscopic Instrument (DESI) Legacy Imaging Surveys Data Release 10 (DR10). This paper is the fourth in a series of searches. This is the first catalog of lens candidates covering nearly the entirety of the extragalactic sky south of declination δ ≈ +32 ∘ , all observed by DECam, covering ∼14,000 deg 2 . We impose a z -band magnitude cut of <20 in AB magnitude. We deploy a residual neural network and EfficientNet as an ensemble trained on a compilation of known lensing systems and high-grade candidates as well as nonlenses in the same footprint. The predictions from these two base models are aggregated using a meta-learner. After applying our ensemble to the survey data, we exclude known candidates and systems, and use our own visual inspection portal to rank images in the top 0.01 percentile of all neural network recommendations. We have found 811 lens candidates, five of which are confirmed through Euclid Quick Data Release (Q1). These include 484 new candidates in the Legacy Surveys DR9 footprint, all parts of which have been searched for strong lenses at least once before, either by our group or others. Combining the discoveries from this work with those from the first three papers in this series (335, 1210, and 1512), we have discovered a total of 3868 new candidates in the DESI Legacy Surveys.

Inchausti, Jose Carlos [University of San Francisc↗
NASA NTRS1991

Fuzzy expert systems vs. neural networks - Truck backer-upper control revisited

It is pointed out that by merging the advantages of fuzzy expert systems and neural networks one can arrive at a more powerful yet more flexible system for inferencing and learning. The advantages of fuzzy expert systems are their ability to provide nonlinear mapping through the membership functions and fuzzy rules, and the ability to deal with fuzzy information and incomplete and/or imprecise data. The merger of these two concepts is explained using the truck backer-upper control problem. Novel network architectures obtained by merging these two concepts and simulation results for the truck backer-upper problem using the architecture are shown.

Ramamoorthy, P. A.↗
DOE OSTI2022

WPGNN and PLayGen (Wind Plant Graph Neural Network and Plant Layout Generator) [SWR-21-90]

WPGNN is the graph neural network machine learning based surrogate model and software that provides a streamlined approximation of wind plant wake models. It can rapidly estimate the energy production of the plant and turbines for any arbitrarily sized wind plant and layout under any inflow condition. Associated tools include graphing and visualization capabilities as well as a wind plant generator capable of creating randomized realizations of canonical wind plant layouts. The WPGNN architecture and application is extendable to multiple wake models, turbine technologies and features, and supports downstream optimizations of wind plant layouts and control strategies. In addition to the WPGNN, we include the code for the plant layout generator (PLayGen) playgen.py. This generator can produce random realizations of realistic wind plant layouts from one of the four canonical styles: cluster, single string, multiple string, or parallel string. The PLayGen_demo.ipynb notebook provides a demonstration of how to use the generator tool.

Harrison-Atlas, Dylan↗
DOE OSTI2024

Unconventional Wells Interference: Supervised Machine Learning for Detecting Fracture Hits

The primary objective of the study was development of a machine learning (ML)-based workflow for fracture hit (“frac hit”) detection and monitoring using shale oil-field data such as drilling surveys, production history (oil and produced water), pressure, and fracking start time and duration records. The ML method takes advantage of long short-term memory (LSTM) and multilayer perceptron (MLP) neural networks to identify the frac hits due to hydraulic communication between the fracking child well(s) and the producing parent well(s) within the same pad (intra-pad interaction) and/or on different pads (inter-pad interaction). It utilizes time series of pressure and production data from within a pad and from adjacent pads. The workflow can capture time variable features of frac hits when the model architecture is deep and wide enough, with enough trainable parameters for deep learning and feature extraction, as demonstrated in this paper by using training and testing subsets of the field data from selected neighboring pads with over a couple of hundred wells. The study was focused on frac-hit interaction among paired wells and demonstrated that the ML model, once trained, can predict the frac-hit probability.

58 GEOSCIENCES↗
DOE OSTI2023

Phonon-informed Neural Thermal Scattering (NeTS) Optimization for Crystalline Graphite and Beryllium Metal

Fast neutrons born from fission lose energy through scattering interactions in the process of slowing-down. As neutrons thermalize to the order of $k$ $b$ $T$ (where $k$ $b$ is the Boltzmann constant, and $T$ is the temperature of the medium), their de Broglie wavelength and energy approaches the order of inter-atomic spacing and quantized lattice vibrations, i.e., phonons. At thermal energies, the thermal scattering law (TSL), i.e., $S$($α, β$), captures crystal binding contributions to the total reaction rate, or cross section. This dimensionless material property describes the energy ($β$) and momentum ($α$) exchanges available in a medium. Currently, $S$($α, β$) is evaluated in the Full Law Analysis Scattering System Hub (FLASSH) code for discrete inputs and stored as ENDF/B File 7 for 0-phonon elastic (MT 2) and n-phonon inelastic (MT 4) processes. Further processing recasts $S$($α, β$) into cumulative distribution functions for sampling post-collision scattering kinematics. In practice, interpolation schemes are employed to access data between tabulated values. An improvement to this juncture of the nuclear data pipeline is supplying cross sections on-the-fly (OTF), as has been developed for the un-resolved resonance region to minimize non-physical interpolation errors. This capability may improve simulation accuracy for accident and transient analyses, where rapidly varying changes in temperature and pressure are difficult to predict beforehand. To do so, deep artificial neural networks (ANNs) can be employed which collapse non-linear, complex data into a lightweight dictionary of neural weights and biases. This has been successfully demonstrated for the hydrogen in light water $S$($α, β$) dataset in the form of a Neural Thermal Scattering (NeTS) module. In this work, the NeTS framework is extended to consider the impact of material-dependent dynamical features on optimal neural pre-processing and architecture design decisions, such as number of neurons per hidden layer, residual skip connections and neural depth. New NeTS modules for crystalline graphite and beryllium metal illuminate a novel correlation between dynamical nonlinearity and optimal neural parametrization when deploying $S$($α, β$) on-the-fly.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
NASA NTRSDate not supplied

Modeling Deformable Linear Objects for Autonomous Robotic Outfitting of Lunar Surface Systems

This paper presents structural models of deformable linear objects (DLOs). DLOs are a subclass of deformable objects that encompasses common outfitting elements such as cables and ropes. Models are validated through hardware experiments, and integration in a robotic autonomy architecture for space environments is discussed. A persistent human presence on the lunar surface is one of the next major milestones in space exploration. This requires the development of robust extraplanetary construction technologies including structures and materials modeling and robotic systems. Previous robotic construction technology development has primarily focused on structural assembly, with significantly less focus on robotically performed outfitting tasks to instantiate subsystems providing power, data, life support, etc. These tasks involve manipulation of highly flexible elements, which are difficult to model, such as cable harnesses, ropes, and hoses. Robotic manipulation of DLOs, especially cable harnesses, is an active area of research as cable harnesses are essential for providing power and data to space assets. DLO models that can be used for robot manipulator trajectory generation are necessary for autonomous operation of lunar infrastructure. There are many proposed methods for modeling DLOs, and they primarily fall into three types: 1) discrete model-based, 2) continuum model-based, and 3) Neural Network-based. These types each have pros and cons, and the tradeoff between model accuracy and computational speed informs which type should be used. An understanding of this trade-off is imperative for real-time control of autonomous systems. High computational requirements reduce the speed of the model, making real-time control difficult, while accuracy is critical to preventing collisions. Discrete models, such as a mass-spring multibody representation, require relatively few calculations, and accuracy is directly tied to the step size of the discretization. Continuum models, such as a B-spline representation or a Cosserat rod model (a mix of continuous and discrete), are more informed of the structural properties of the cable and are much more accurate than a rigid body mass-spring model, but at significant computational cost. A Neural Network approach can provide an online solution with very few computational steps, but properly generating training data can be difficult and validation for an in-space application is not trivial. This paper explores the trade-off between different modeling approaches and compares accuracy and computational speed/complexity of the three types mentioned above. Model accuracy is evaluated using a cable in a static configuration. True cable shape is obtained using a depth camera for RGB images and point-cloud segmentation. The purpose of this experiment is to evaluate the trade-offs of different approaches to the DLO modeling problem. Understanding the tradeoffs between different cable modeling techniques paves the way for developing robotic control and planning architectures necessary for real-time manipulation of DLOs for lunar infrastructure outfitting. Real-time control is required for robotic systems to be able to actively manipulate a cable in a harsh environment where model and sensor errors compound, and environmental conditions can cause significant disturbances. Cable routing must be performed in areas with high density of objects/obstacles: through truss structures, near solar panels or mirror arrays, next to bundles of electrical equipment. Understanding the best way to plan and manipulate a cable without disrupting the environment or damaging the cable is imperative to robotic outfitting operations on the lunar surface.

Amy M Quartaro↗
DOE OSTI2023

Learning macroscopic internal variables and history dependence from microscopic models

This paper concerns the study of history dependent phenomena in heterogeneous materials in a two-scale setting where the material is specified at a fine microscopic scale of heterogeneities that is much smaller than the coarse macroscopic scale of application. Here, we specifically study a polycrystalline medium where each grain is governed by crystal plasticity while the solid is subjected to macroscopic dynamic loads. The theory of homogenization allows us to solve the macroscale problem directly with a constitutive relation that is defined implicitly by the solution of the microscale problem. However, the homogenization leads to a highly complex history dependence at the macroscale, one that can be quite different from that at the microscale. In this paper, we examine the use of machine-learning, and especially deep neural networks, to harness data generated by repeatedly solving the finer scale model to: (i) gain insights into the history dependence and the macroscopic internal variables that govern the overall response; and (ii) to create a computationally efficient surrogate of its solution operator, that can directly be used at the coarser scale with no further modeling. We do so by introducing a recurrent neural operator (RNO), and show that: (i) the architecture and the learned internal variables can provide insight into the physics of the macroscopic problem; and (ii) that the RNO can provide multiscale, specifically FE 2 , accuracy at a cost comparable to a conventional empirical constitutive relation.

36 MATERIALS SCIENCE↗
NASA NTRS1987

Learning in Artificial Neural Systems

This paper presents an overview and analysis of learning in Artificial Neural Systems (ANS's). It begins with a general introduction to neural networks and connectionist approaches to information processing. The basis for learning in ANS's is then described, and compared with classical Machine learning. While similar in some ways, ANS learning deviates from tradition in its dependence on the modification of individual weights to bring about changes in a knowledge representation distributed across connections in a network. This unique form of learning is analyzed from two aspects: the selection of an appropriate network architecture for representing the problem, and the choice of a suitable learning rule capable of reproducing the desired function within the given network. The various network architectures are classified, and then identified with explicit restrictions on the types of functions they are capable of representing. The learning rules, i.e., algorithms that specify how the network weights are modified, are similarly taxonomized, and where possible, the limitations inherent to specific classes of rules are outlined.

Matheus, Christopher J.↗
DOE OSTI2026

Community detection robustness of graph neural networks

Graph neural networks (GNNs) are increasingly widely used for community detection in attributed networks. They combine structural topology with node attributes through message passing and pooling. However, their robustness or lack thereof with respect to different perturbations and targeted attacks in conjunction with community detection tasks is not well understood. To shed light on latent mechanisms behind GNN sensitivity on community detection tasks, we conduct a systematic computational evaluation of six widely adopted GNN architectures graph convolutional network, graph attention network, graph sample and aggregate (GraphSAGE), differentiable pooling (DiffPool), minimum cut pooling (MinCUT), and deep modularity networks (DMoN). The analysis covers three perturbation categories: node attribute manipulations, edge topology distortions, and adversarial attacks. We use element-centric similarity as the evaluation metric on synthetic benchmarks and real-world citation networks. Our findings indicate that supervised GNNs tend to achieve higher baseline accuracy, while unsupervised methods, particularly DMoN, maintain stronger resilience under targeted and adversarial perturbations. Furthermore, robustness appears to be strongly influenced by community strength, with well-defined communities reducing performance loss. Across all models, node attribute perturbations associated with targeted edge deletions and shifts in attribute distributions tend to cause the largest degradation in community recovery. These findings highlight important trade-offs between accuracy and robustness in GNN-based community detection and offer insights into selecting architectures resilient to noise and adversarial attacks.

Goel, Jaidev [Virginia Polytechnic Inst. and State↗
DOE OSTI2022

Vector-Matrix Multiplication Engine for Neuromorphic Computation with a CBRAM Crossbar Array [Slides]

The core function of many neural network algorithms is the dot product, or vector matrix multiply (VMM) operation. Crossbar arrays utilizing resistive memory elements can reduce computational energy in neural algorithms by up to five orders of magnitude compared to conventional CPUs. Moving data between a processor, SRAM, and DRAM dominates energy consumption. By utilizing analog operations to reduce data movement, resistive memory crossbars can enable processing of large amounts of data at lower energy than conventional memory architectures.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Analyzing inference workloads for spatiotemporal modeling

Ensuring power grid resiliency, forecasting climate conditions, and optimization of transportation infrastructure are some of the many application areas where data is collected in both space and time. Spatiotemporal modeling is about modeling those patterns for forecasting future trends and carrying out critical decision-making by leveraging machine learning/deep learning. Once trained offline, field deployment of trained models for near real-time inference could be challenging because performance can vary significantly depending on the environment, available compute resources and tolerance to ambiguity in results. Users deploying spatiotemporal models for solving complex problems can benefit from analytical studies considering a plethora of system adaptations to understand the associated performance-quality trade-offs. To facilitate the co-design of next-generation hardware architectures for field deployment of trained models, it is critical to characterize the workloads of these deep learning (DL) applications during inference and assess their computational patterns at different levels of the execution stack. In this paper, we develop several variants of deep learning applications that use spatiotemporal data from dynamical systems. We study the associated computational patterns for inference workloads at different levels, considering relevant models (Long short-term Memory, Convolutional Neural Network and Spatio-Temporal Graph Convolution Network), DL frameworks (Tensorflow and PyTorch), precision (FP16, FP32, AMP, INT16 and INT8), inference runtime (ONNX and AI Template), post-training quantization (TensorRT) and platforms (Nvidia DGX A100 and Sambanova SN10 RDU). Overall, our findings indicate that although there is potential in mixed-precision models and post-training quantization for spatiotemporal modeling, extracting efficiency from contemporary GPU systems might be challenging. Instead, co-designing custom accelerators by leveraging optimized High Level Synthesis frameworks (such as SODA High-Level Synthesizer for customized FPGA/ASIC targets) can make workload-specific adjustments to enhance the efficiency.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Automated Generation of Integrated Digital and Spiking Neuromorphic Machine Learning Accelerators

The growing numbers of application areas for artificial intelligence (AI) methods have led to an explosion of domain-specific accelerators that could support every new machine learning (ML) algorithm advancement, clearly highlighting the need for a capability to quickly and automatically transition from algorithm definition to hardware implementation and explore design space along a variety of SWaP (size, weight and Power). The software defined architectures (SODA) synthesizer implements a compiler-based modular infrastructure for the end-to-end generation of machine learning accelerators from high-level frameworks to hardware description language. At the same time, neuromorphic computing, by mimicking how the brain operates, promises to perform artificial intelligence tasks at efficiencies orders of magnitude higher than the current conventional tensor-processing based accelerators, as demonstrated by a variety of specialized designs leveraging Spiking Neural Networks (SNNs). Nevertheless, the mapping of an artificial neural network (ANN) to solutions supporting SNNs is still a non-trivial and very device-specific task, and completely lack the possibility to design hybrid systems that integrate conventional and spiking neural models. In this paper we discuss the support for such an integrated generation leveraging the SODA Synthesizer framework and its modular structure. In particular, we present a new MLIR dialect (part of the SODA frontend) that allows expressing spiking neural network features (e.g., available resources, spiking sequences, analog signal reading, etc.) and illustrate how it enables mapping to Spiking Neurons and deployment to the related specialized hardware (which, in the digital domain, could be generated through the other existing layers of the SODA Synthesizer). We then discuss the opportunities for even deeper integration afforded by the hardware compilation infrastructure, providing a path towards the generation of complex heterogeneous artificial intelligence systems.

Curzel, Serena↗
NASA NTRS2005

Hot Chips and Hot Interconnects for High End Computing Systems

I will discuss several processors: 1. The Cray proprietary processor used in the Cray X1; 2. The IBM Power 3 and Power 4 used in an IBM SP 3 and IBM SP 4 systems; 3. The Intel Itanium and Xeon, used in the SGI Altix systems and clusters respectively; 4. IBM System-on-a-Chip used in IBM BlueGene/L; 5. HP Alpha EV68 processor used in DOE ASCI Q cluster; 6. SPARC64 V processor, which is used in the Fujitsu PRIMEPOWER HPC2500; 7. An NEC proprietary processor, which is used in NEC SX-6/7; 8. Power 4+ processor, which is used in Hitachi SR11000; 9. NEC proprietary processor, which is used in Earth Simulator. The IBM POWER5 and Red Storm Computing Systems will also be discussed. The architectures of these processors will first be presented, followed by interconnection networks and a description of high-end computer systems based on these processors and networks. The performance of various hardware/programming model combinations will then be compared, based on latest NAS Parallel Benchmark results (MPI, OpenMP/HPF and hybrid (MPI + OpenMP). The tutorial will conclude with a discussion of general trends in the field of high performance computing, (quantum computing, DNA computing, cellular engineering, and neural networks).

Saini, Subhash↗
DOE OSTI2021

Few-shot Learning for Post-disaster Structure Damage Assessment

Automating post-disaster damage assessment with remote sensing data is critical for faster surveys of structures impacted by natural disasters. One significant obstacle to training state-of-the-art deep neural networks to support this automation is that large quantities of labelled data are often required. However, obtaining those labels is particularly unrealistic to support post-disaster damage assessment in a timely manner. Few-shot learning methods could help to mitigate this by reducing the amount of labelled data required to successfully train a model while achieving satisfactory results. To this end, we explore a feature reweighting method to the YOLOv3 object detection architecture to achieve few-shot learning of damage assessment models on the xBD dataset. Our results show that the feature reweighting approach yield improved mAP over the baseline with significantly fewer labelled samples. In addition, we use t-SNE to analyze the class-specific reweighting vectors generated by the reweighting module in order to evaluate their inter-class and intra-class similarity. We find that the vectors form clusters based on class, and that these clusters overlap with visually similar classes. Those results show the potential to employ this few-shot learning strategy for rapid damage assessment with post-event remote sensing images.

Bowman, Jordan↗