Search NASA⌕ Search

SEARCH · Search NASA

Results for “hardware neural networks”

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 145 records · Page 8
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Low Power Hardware-In-The-Loop Neuromorphic Training Accelerator

The training process for spiking neural networks can be very computationally intensive. Approaches such as evolutionary algorithms may require evaluating thousands or millions of candidate solutions. In this work, we propose using neuromorphic cores implemented on a Xilinx Zynq system on chip to accelerate and improve the energy efficiency of the evaluation step of an evolutionary training approach. We demonstrate this can significantly reduce the required energy to evolve a network with some cases showing greater than 10 times improvement as compared to a CPU-only system.

Mitchell, Parker↗
DOE OSTI2021

Binary Complex Neural Network Acceleration on FPGA

Being able to learn from complex data with phase information is imperative for many signal processing applications. Today’s real-valued deep neural networks (DNNs) have shown efficiency in latent information analysis but fall short when applied to the complex domain. Deep complex networks (DCN) , in contrast, can learn from complex data, but have high computational costs; therefore, they cannot satisfy the instant decision making requirements of many deployable systems dealing with short observations or short signal bursts. Recent, Binarized Complex Neural Network (BCNN), which integrates DCNs with binarized neural networks (BNN), shows great potential in classifying complex data in real-time. In this paper, we propose a structural pruning based accelerator of BCNN, which is able to provide more than 5000 frames/s inference throughput on edge devices. The high performance comes from both the algorithm and hardware sides. On the algorithm side, we conduct structural pruning to the original BCNN models and obtain 20 × pruning rates with negligible accuracy loss; on the hardware side, we propose a novel 2D convolution operation accelerator for the binary complex neural network. Experimental results show that the proposed design works with over 90% utilization and is able to achieve the inference throughput of 5882 frames/s and 4938 frames/s for complex NIN-Net and ResNet-18 using CIFAR-10 dataset and Alveo U280 Board.

Peng, Hongwu↗
DOE OSTI2024

Efficient Mixed-Precision Matrix Factorization of the Inverse Overlap Matrix in Electronic Structure Calculations with AI-Hardware and GPUs

In recent years, a new kind of accelerated hardware has gained popularity in the artificial intelligence (AI) community which enables extremely high-performance tensor contractions in reduced precision for deep neural network calculations. In this article, we exploit Nvidia Tensor cores, a prototypical example of such AI-hardware, to develop a mixed precision approach for computing a dense matrix factorization of the inverse overlap matrix in electronic structure theory, S –1 . This factorization of S –1 , written as ZZT = S –1 , is used to transform the general matrix eigenvalue problem into a standard matrix eigenvalue problem. Here we present a mixed precision iterative refinement algorithm where Z is given recursively using matrix–matrix multiplications and can be computed with high performance on Tensor cores. To understand the performance and accuracy of Tensor cores, comparisons are made to GPU-only implementations in single and double precision. Additionally, we propose a nonparametric stopping criteria which is robust in the face of lower precision floating point operations. The algorithm is particularly useful when we have a good initial guess to Z, for example, from previous time steps in quantum-mechanical molecular dynamics simulations or from a previous iteration in a geometry optimization.

36 MATERIALS SCIENCE↗
DOE OSTI2024

Scaling neural simulations in STACS

Abstract As modern neuroscience tools acquire more details about the brain, the need to move towards biological-scale neural simulations continues to grow. However, effective simulations at scale remain a challenge. Beyond just the tooling required to enable parallel execution, there is also the unique structure of the synaptic interconnectivity, which is globally sparse but has relatively high connection density and non-local interactions per neuron. There are also various practicalities to consider in high performance computing applications, such as the need for serializing neural networks to support potentially long-running simulations that require checkpoint-restart. Although acceleration on neuromorphic hardware is also a possibility, development in this space can be difficult as hardware support tends to vary between platforms and software support for larger scale models also tends to be limited. In this paper, we focus our attention on Simulation Tool for Asynchronous Cortical Streams (STACS), a spiking neural network simulator that leverages the Charm++ parallel programming framework, with the goal of supporting biological-scale simulations as well as interoperability between platforms. Central to these goals is the implementation of scalable data structures suitable for efficiently distributing a network across parallel partitions. Here, we discuss a straightforward extension of a parallel data format with a history of use in graph partitioners, which also serves as a portable intermediate representation for different neuromorphic backends. We perform scaling studies on the Summit supercomputer, examining the capabilities of STACS in terms of network build and storage, partitioning, and execution. We highlight how a suitably partitioned, spatially dependent synaptic structure introduces a communication workload well-suited to the multicast communication supported by Charm++. We evaluate the strong and weak scaling behavior for networks on the order of millions of neurons and billions of synapses, and show that STACS achieves competitive levels of parallel efficiency.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Hydrogen-induced tunable remanent polarization in a perovskite nickelate

Materials with field-tunable polarization are of broad interest to condensed matter sciences and solid-state device technologies. Here, using hydrogen (H) donor doping, we modify the room temperature metallic phase of a perovskite nickelate NdNiO 3 into an insulating phase with both metastable dipolar polarization and space-charge polarization. We then demonstrate transient negative differential capacitance in thin film capacitors. The space-charge polarization caused by long-range movement and trapping of protons dominates when the electric field exceeds the threshold value. First-principles calculations suggest the polarization originates from the polar structure created by H doping. We find that polarization decays within ~1 second which is an interesting temporal regime for neuromorphic computing hardware design, and we implement the transient characteristics in a neural network to demonstrate unsupervised learning. These discoveries open new avenues for designing ferroelectric materials and electrets using light-ion doping.

36 MATERIALS SCIENCE↗
NASA NTRS1989

Computational chaos in massively parallel neural networks

A fundamental issue which directly impacts the scalability of current theoretical neural network models to massively parallel embodiments, in both software as well as hardware, is the inherent and unavoidable concurrent asynchronicity of emerging fine-grained computational ensembles and the possible emergence of chaotic manifestations. Previous analyses attributed dynamical instability to the topology of the interconnection matrix, to parasitic components or to propagation delays. However, researchers have observed the existence of emergent computational chaos in a concurrently asynchronous framework, independent of the network topology. Researcher present a methodology enabling the effective asynchronous operation of large-scale neural networks. Necessary and sufficient conditions guaranteeing concurrent asynchronous convergence are established in terms of contracting operators. Lyapunov exponents are computed formally to characterize the underlying nonlinear dynamics. Simulation results are presented to illustrate network convergence to the correct results, even in the presence of large delays.

Barhen, Jacob↗
DOE OSTI2026

Measuring the Energy Consumption and Efficiency of Deep Neural Networks: An Empirical Analysis and Design Recommendations

Addressing the "Red-AI" trend of rising energy consumption by large-scale neural networks, this study investigates the measured energy consumption of training various fully connected neural network architectures. We introduce the BUTTER-E dataset, an augmentation to the BUTTER Empirical Deep Learning dataset, containing energy consumption and performance data from 41,129 individual experimental runs spanning 30,582 distinct configurations: 13 datasets, 20 sizes (trainable parameters), 8 "shapes", and 14 depths on both CPUs and GPUs using node-level watt-meters. This dataset reveals the complex relationship between dataset size, network structure, and energy use. Our analysis uncovers a surprising, hardware-mediated non-linear relationship between energy efficiency and network design, challenging the assumption that reducing the number of parameters or FLOPs is the best way to achieve greater energy efficiency. We propose a straightforward and effective energy model that accounts for network size, computing, and memory hierarchy. Highlighting the need for cache-considerate algorithm development, we suggest a codesign approach to energy efficient network, algorithm, and hardware design. This work contributes to the fields of sustainable computing and Green AI, offering practical guidance for creating more energy-efficient neural networks and promoting sustainable AI.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2006

Dynamics and Adaptive Control for Stability Recovery of Damaged Aircraft

This paper presents a recent study of a damaged generic transport model as part of a NASA research project to investigate adaptive control methods for stability recovery of damaged aircraft operating in off-nominal flight conditions under damage and or failures. Aerodynamic modeling of damage effects is performed using an aerodynamic code to assess changes in the stability and control derivatives of a generic transport aircraft. Certain types of damage such as damage to one of the wings or horizontal stabilizers can cause the aircraft to become asymmetric, thus resulting in a coupling between the longitudinal and lateral motions. Flight dynamics for a general asymmetric aircraft is derived to account for changes in the center of gravity that can compromise the stability of the damaged aircraft. An iterative trim analysis for the translational motion is developed to refine the trim procedure by accounting for the effects of the control surface deflection. A hybrid direct-indirect neural network, adaptive flight control is proposed as an adaptive law for stabilizing the rotational motion of the damaged aircraft. The indirect adaptation is designed to estimate the plant dynamics of the damaged aircraft in conjunction with the direct adaptation that computes the control augmentation. Two approaches are presented 1) an adaptive law derived from the Lyapunov stability theory to ensure that the signals are bounded, and 2) a recursive least-square method for parameter identification. A hardware-in-the-loop simulation is conducted and demonstrates the effectiveness of the direct neural network adaptive flight control in the stability recovery of the damaged aircraft. A preliminary simulation of the hybrid adaptive flight control has been performed and initial data have shown the effectiveness of the proposed hybrid approach. Future work will include further investigations and high-fidelity simulations of the proposed hybrid adaptive Bight control approach.

Nguyen, Nhan↗
DOE OSTI2025

Enabling Real-Time Communication in Multi-Agent Systems: A Graph Neural Network Based Approach

Global connectivity enables effective coordination in Multi-Agent Systems (MAS). Solving these connection problems under hardware constraints is an NP-hard non-Euclidean Degree Constrained Minimum Spanning Tree (DCMST) problem. Prior MAS controllers coordinate team movement for task completion and collision avoidance; some considering Line-of-Sight (LOS) maintenance but prioritizing flexibility over guarantees. Evolutionary Algorithms (EA) have been shown to find good solutions for DCMST, but their performance degrades with larger populations required to support a large MAS. We present a method based on edge graph attention networks, trained offline to reduce online computation times. Empirical comparisons with greedy polynomial-time solvers and EA show that our method leverages latent graph information to consistently find constraint-satisfying solutions in less time.

connectivity maintenance↗
DOE OSTI2023

Selective area doping for Mott neuromorphic electronics

The cointegration of artificial neuronal and synaptic devices with homotypic materials and structures can greatly simplify the fabrication of neuromorphic hardware. We demonstrate experimental realization of vanadium dioxide (VO 2 ) artificial neurons and synapses on the same substrate through selective area carrier doping. By locally configuring pairs of catalytic and inert electrodes that enable nanoscale control over carrier density, volatility or nonvolatility can be appropriately assigned to each two-terminal Mott memory device per lithographic design, and both neuron- and synapse-like devices are successfully integrated on a single chip. Feedforward excitation and inhibition neural motifs are demonstrated at hardware level, followed by simulation of network-level handwritten digit and fashion product recognition tasks with experimental characteristics. Spatially selective electron doping opens up previously unidentified avenues for integration of emerging correlated semiconductors in electronic device technologies.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2023

Bayesian quantum state reconstruction with a learning-based tuned prior

We demonstrate machine-learning-enhanced Bayesian quantum state tomography on near-term intermediate-scale quantum hardware. Our approach to selecting prior distributions leverages pre-trained neural networks incorporating measurement data and enables improved inference times over standard prior distributions.

Regmi, Sangita↗
NASA NTRS2019

Deep Learning Method for Detecting Precursors to Adverse Events

With the recent advancements in Deep Learning methods, the ability to model large complex heterogeneous data sets are fundamentally changing industry and research. Coupled with hardware improvements, and ease of implementation, a wide variety of deep neural network architectures can quickly be developed to solve a sweeping range of problems such as: object detection in images, automatic healthcare diagnosis using heterogenous data sources, real time language translating and sentence prediction, upscaling low resolution images, and forecasting of multivariate timeseries. Generally, many of these architectures outperform classical machine learning approaches in their respective tasks, however, this typically comes at a cost of interpretability. These black box algorithms generally suffer from lack of transparency in both model complexity as well as the rationale behind the prediction. This lack of comprehension, is driving an emerging area of interest in “Explainable AI”. An algorithm called: “Deep Temporal Multiple Instance Learning”1 was a recently developed to identify precursors to adverse events and has been applied in the aviation domain. The deep learning architecture is designed to capture the evolution of the probability of the outcome over the time preceding the adverse event using a multiple instance learning approach as illustrated in Figure 1. Precursors are defined when the probability of the event has exceeded a threshold at some point in the timeseries, at which point, a sensitivity analysis is performed to determine contributing factors. The contributing factors are used to explain and define the precursor during the periods where the probability score is high. The identified contributing factors are then presented to subject matter experts to provide objective insights into the leading factors associated with the particular adverse event. The algorithm has been tested on flight data from a commercial airline and has the ability to discover precursors to known adverse events that take the form of safety critical operations, such as unstable approach events on final approach. Apart from detecting precursors to adverse events, the converse can also be leveraged to discover corrective actions. These positive actions manifest themselves as periods in the timeseries when the precursor score has been lowered from an elevated state; meaning that if the system had been left uncorrected, it would have eventually reached the adverse event state. Characterizing these state changes can help identify successful interventions that may not have been known before. Policy makers and procedure designers can use this additional knowledge to craft more safety and efficient resilient procedures for future operations and therefore improve the overall performance of the National Airspace.

Matthews, Bryan L.↗
DOE OSTI2023

Application of Quantum Machine Learning to High Energy Physics Analysis at LHC using IBM Quantum Computer Simulators and IBM Quantum Computer Hardware

Our group pioneers the use of Quantum Machine Learning (QML) on High Energy Physics analysis at LHC. We have successfully employed several QML classification algorithms in the ttH (Higgs production in association with a top quark pair) and Higgs to two muons (Higgs coupling to second generation fermions), two recent LHC flagship physics analysis, on gate-model quantum computer simulators and hardware. The simulation studies have been performed with the IBM Quantum Framework, Google Tensorflow Quantum Framework, and Amazon Braket Framework, and we have achieved good classification performance that is similar to the performances of the classical machine learning methods currently used in LHC physics analyses, classical SVM, classical BDT, and classical deep neural network for example. We have also performed our studies using IBM superconducting quantum computer hardware and the performance is promising and is approaching the performance from IBM quantum simulators. Moreover, we extend our studies to other QML areas such as quantum anomaly detection and quantum generative adversarial, and some preliminary results have been obtained. Also, we have overcome the challenges of intensive computing resources in the cases of large qubits (25 qubits or more) and large numbers of events using NVIDIA cuQuantum with NERSC Perlmutter HPC. Our studies give an example that Quantum Machine Learning performs as well as its classical counterpart for realistic High Energy Physics analysis datasets. Furthermore, our result on noisy quantum hardware provides important validation for the result on noiseless quantum simulators.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2021

A Software Framework for Comparing Training Approaches for Spiking Neuromorphic Systems

There are a wide variety of training approaches for spiking neural networks for neuromorphic deployment. However, it is often not clear how these training algorithms perform or compare when applied across multiple neuromorphic hardware platforms and multiple datasets. In this work, we present a software framework for comparing performance across four neuromorphic training algorithms across three neuromorphic simulators and four simple classification tasks. We introduce an approach for training a spiking neural network using a decision tree, and we compare this approach to training algorithms based on evolutionary algorithms, back-propagation, and reservoir computing. We present a hyperparameter optimization approach to tune the hyperparameters of the algorithm, and show that these optimized hyperparameters depend on the processor, algorithm, and classification task. Finally, we compare the performance of the optimized algorithms across multiple metrics, including accuracy, training time, and resulting network size, and we show that there is not one best training algorithm across all datasets and performance metrics.

Schuman, Catherine↗
NASA NTRS2005

Creation and testing of an artificial neural network based carbonate detector for Mars rovers

We have developed an artificial neural network (ANN) based carbonate detector capable of running on current and future rover hardware. The detector can identify calcite in visible/NIR (350-2500 nm) spectra of both laboratory specimens covered by ferric dust and rocks in Mars analogue field environments. The ANN was trained using the Backpropagation algorithm with sigmoid activation neurons. For the training dataset, we chose nine carbonate and eight non-carbonate representative mineral spectra from the USGS spectral library. Using these spectra as seeds, we generated 10,000 variants with up to 2% Gaussian noise in each reflectance measurement. We cross-validated several ANN architectures, training on 9,900 spectra and testing on the remaining 100. The best performing ANN correctly detected, with perfect accuracy, the presence (or absence) of carbonate in spectral data taken on field samples from the Mojave desert and clean, pure marbles from CT. Sensitivity experiments with JSC Mars-1 simulant dust suggest the carbonate detector would perform well in aeolian Martian environments.

neural networks↗
DOE OSTI2024

Machine learning without a processor: Emergent learning in a nonlinear analog network

Standard deep learning algorithms require differentiating large nonlinear networks, a process that is slow and power-hungry. Electronic contrastive local learning networks (CLLNs) offer potentially fast, efficient, and fault-tolerant hardware for analog machine learning, but existing implementations are linear, severely limiting their capabilities. These systems differ significantly from artificial neural networks as well as the brain, so the feasibility and utility of incorporating nonlinear elements have not been explored. Here, we introduce a nonlinear CLLN—an analog electronic network made of self-adjusting nonlinear resistive elements based on transistors. We demonstrate that the system learns tasks unachievable in linear systems, including XOR (exclusive or) and nonlinear regression, without a computer. We find our decentralized system reduces modes of training error in order (mean, slope, curvature), similar to spectral bias in artificial neural networks. The circuitry is robust to damage, retrainable in seconds, and performs learned tasks in microseconds while dissipating only picojoules of energy across each transistor. This suggests enormous potential for fast, low-power computing in edge systems like sensors, robotic controllers, and medical devices, as well as manufacturability at scale for performing and studying emergent learning.

Science & Technology - Other Topics↗
DOE OSTI2020

PANNA: Properties from Artificial Neural Network Architectures

We report prediction of material properties from first principles is often a computationally expensive task. Recently, artificial neural networks and other machine learning approaches have been successfully employed to obtain accurate models at a low computational cost by leveraging existing example data. Here, we present a software package “Properties from Artificial Neural Network Architectures” (PANNA) that provides a comprehensive toolkit for creating neural network models for atomistic systems following the Behler–Parrinello topology. Besides the core routines for neural network training, it includes data parser, descriptor builder for Behler–Parrinello class of symmetry functions and force-field generator suitable for integration within molecular dynamics packages. PANNA offers a variety of activation and cost functions, regularization methods, as well as the possibility of using fully-connected networks with custom size for each atomic species. PANNA benefits from the optimization and hardware-flexibility of the underlying TensorFlow engine which allows it to be used on multiple CPU/GPU/TPU systems, making it possible to develop and optimize neural network models based on large datasets.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
NASA NTRS2009

Implementation of an Adaptive Controller System from Concept to Flight Test

The National Aeronautics and Space Administration Dryden Flight Research Center (Edwards, California) is conducting ongoing flight research using adaptive controller algorithms. A highly modified McDonnell-Douglas NF-15B airplane called the F-15 Intelligent Flight Control System (IFCS) is used to test and develop these algorithms. Modifications to this airplane include adding canards and changing the flight control systems to interface a single-string research controller processor for neural network algorithms. Research goals include demonstration of revolutionary control approaches that can efficiently optimize aircraft performance in both normal and failure conditions and advancement of neural-network-based flight control technology for new aerospace system designs. This report presents an overview of the processes utilized to develop adaptive controller algorithms during a flight-test program, including a description of initial adaptive controller concepts and a discussion of modeling formulation and performance testing. Design finalization led to integration with the system interfaces, verification of the software, validation of the hardware to the requirements, design of failure detection, development of safety limiters to minimize the effect of erroneous neural network commands, and creation of flight test control room displays to maximize human situational awareness; these are also discussed.

Larson, Richard R.↗