Search NASA⌕ Search

SEARCH · Search NASA

Results for “Machine Learning Hardware”

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 109 records · Page 6
All sourcesNASA NTRSDOE OSTI
DOE OSTI2019

Performance optimization, modeling and analysis of sparse matrix-matrix products on multi-core and many-core processors

Sparse matrix-matrix multiplication (SpGEMM) is a computational primitive that is widely used in areas ranging from traditional numerical applications to recent big data analysis and machine learning. Although many SpGEMM algorithms have been proposed, hardware specific optimizations for multi- and many-core processors are lacking and a detailed analysis of their performance under various use cases and matrices is not available. In this work, we firstly identify and mitigate multiple bottlenecks with memory management and thread scheduling on Intel Xeon Phi (Knights Landing or KNL). Specifically targeting many-core processors, we develop a hash-table-based algorithm and optimize a heap-based shared-memory SpGEMM algorithm. We examine their performance together with other publicly available codes. Different from the literature, our evaluation also includes use cases that are representative of real graph algorithms, such as multi-source breadth-first search or triangle counting. Our hash-table and heap-based algorithms are showing significant speedups from libraries in the majority of the cases while different algorithms dominate the other scenarios with different matrix size, sparsity, compression factor and operation type. We wrap up in-depth evaluation results and make a recipe to give the best SpGEMM algorithm for target scenario. We build the performance model for hash-table and heap-based algorithms, which supports the recipe. A critical finding is that hash-table-based SpGEMM gets a significant performance boost if the nonzeros are not required to be sorted within each row of the output matrix. Finally, we integrate our implementations into a large-scale protein clustering code named HipMCL, accelerating its SpGEMM kernel by up to 10X and achieving an overall performance boost for the whole HipMCL application by 2.6X.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

High-Performance Sparse Matrix-Matrix Products on Intel KNL and Multicore Architectures

Sparse matrix-matrix multiplication (SpGEMM) is a computational primitive that is vastly used in areas ranging from traditional numerical applications to recent big data analysis and machine learning. While many SpGEMM algorithms have been proposed, hardware specific optimizations for multi- and many-core processors are lacking and a detailed analysis of their performance under various use cases and matrices is not available. We firstly identify and mitigate multiple bottlenecks with memory management and thread scheduling on Intel Xeon Phi (Knights Landing or KNL). Specifically targeting multi- and many-core processors, we develop a hash-table-based algorithm and optimize a heap-based shared-memory SpGEMM algorithm. We investigate their performance together with other publicly available codes. Different from the literature, our evaluation also includes use cases that are representative of real graph algorithms, such as multi-source breadth-first search or triangle counting. Our hash-table and heap-based algorithms are showing significant speedups from libraries in the majority of the cases while different algorithms dominate the other scenarios with different matrix size, sparsity, compression factor and operation type. We wrap up in-depth evaluation results and make a recipe to give the best SpGEMM algorithm for target scenario. In conclusion, a critical finding is that hash-table-based SpGEMM gets a significant performance boost if the nonzeros are not required to be sorted within each row of the output matrix.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

AI to Automate ModEx for Optimal Predictive Improvement and Scientific Discovery

Focal Areas: Data acquisition and assimilation enabled by machine learning, AI, advanced experimental optimization, unsupervised learning, and hardware-related AI efforts; Predictive modeling through AI techniques and AI-derived model components; Using AI to design a hierarchical model prediction system consisting and model selection; and, Interrogating complex data (observed and simulated) using AI, big data analytics, and other advanced methods such as explainable AI and physics- or knowledge-guided AI.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

HD-Bind: Encoding of Molecular Structure with Low Precision, Hyperdimensional Binary Representations

Publicly available collections of drug-like molecules have grown to comprise tens of billions of compounds due to advances in combinatorial chemistry. Traditional methods for identifying "hit" molecules from a large collection of potential drug-like candidates have relied on biophysical theory to compute approximations to the Gibbs free energy of the binding interaction between the drug and its protein target. These approaches have the major drawback that they require exceptional computing capabilities for even relatively small collections of molecules. Hyperdimensional Computing (HDC) is a recently-proposed learning paradigm that represents data with high-dimension binary vectors; this allows the use of low-precision binary vector arithmetic to create models of the data that can be learned without the need for the gradient-based optimization required in many conventional machine learning and deep learning methods. This algorithmic simplicity allows for acceleration in hardware that has been previously demonstrated in a range of application areas. We consider existing HDC approaches for molecular property classification and introduce two novel encodings of a commonly-used molecular representation, the extended connectivity fingerprint (ECFP). We show that HDC-based inference methods are as much as 91 times more efficient than traditional machine learning methods, and achieve an acceleration of nearly nine orders of magnitude compared to molecular docking. Our results show that HDC accelerated methods retain competitive accuracy on a number of well-studied tasks such as molecular property predictions using the MoleculeNet dataset, and bind/no-bind activity classification using the DUD-E and LIT-PCBA datasets. Our work thus motivates further investigation into molecular representation learning to develop ultraefficient pre-screening tools.

Jones, William↗
DOE OSTI2021

A review of machine learning in building load prediction

The surge of machine learning in recent years has been empowering engineer modeling in various fields. The decreasing hardware cost, increasing data accessibility, and advances of building automation system (BAS) allow the collection and storage of a significant amount of building operation data. The two facts provide great opportunities of applying machine learning to building energy systems modeling and analysis. There are a great number of research papers on this topic but there lacks a comprehensive and general review to summarize the current development, limitations, gaps and future trend. In this review paper series, machine learning techniques in building energy system modeling and analysis are reviewed under the organization and logic of the machine learning definition by Tom M. Mitchell: a computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E. This paper is the first part of the review paper series, which focuses on building load prediction. First, the applications of building load prediction model (task T) are reviewed. Then, the modeling algorithms improving machine learning performance and accuracy (performance P) are reviewed. At the same time, the literature on the data perspective for modeling (experience E), including data engineering from sensors level to data level, pre-processing, feature extraction and selection, is reviewed. Finally, what is well-studied and what is lacking but with great potential are concluded; the gaps between present and future utilization of machine learning techniques are identified; the future trend and development are also predicted. The target readers of this paper are not only researchers from the building side who can get exposed to cutting edge machine learning tools, but also those from machine learning side who can understand the potential and challenge to apply machine learning in buildings.

Liang, Zhang↗
DOE OSTI2019

FPGA-Accelerated Machine Learning Inference as a Service for Particle Physics Computing

Large-scale particle physics experiments face challenging demands for high-throughput computing resources both now and in the future. New heterogeneous computing paradigms on dedicated hardware with increased parallelization, such as Field Programmable Gate Arrays (FPGAs), offer exciting solutions with large potential gains. The growing applications of machine learning algorithms in particle physics for simulation, reconstruction, and analysis are naturally deployed on such platforms. We demonstrate that the acceleration of machine learning inference as a web service represents a heterogeneous computing solution for particle physics experiments that potentially requires minimal modification to the current computing model. As examples, we retrain the ResNet-50 convolutional neural network to demonstrate state-of-the-art performance for top quark jet tagging at the LHC and apply a ResNet-50 model with transfer learning for neutrino event classification. Using Project Brainwave by Microsoft to accelerate the ResNet-50 image classification model, we achieve average inference times of 60 (10) ms with our experimental physics software framework using Brainwave as a cloud (edge or on-premises) service, representing an improvement by a factor of approximately 30 (175) in model inference latency over traditional CPU inference in current experimental hardware. A single FPGA service accessed by many CPUs achieves a throughput of 600–700 inferences per second using an image batch of one, comparable to large batch-size GPU throughput and significantly better than small batch-size GPU throughput. Deployed as an edge or cloud service for the particle physics computing model, coprocessor accelerators can have a higher duty cycle and are potentially much more cost-effective.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2021

A domain wall-magnetic tunnel junction artificial synapse with notched geometry for accurate and efficient training of deep neural networks

Inspired by the parallelism and efficiency of the brain, several candidates for artificial synapse devices have been developed for neuromorphic computing, yet a nonlinear and asymmetric synaptic response curve precludes their use for backpropagation, the foundation of modern supervised learning. Spintronic devices—which benefit from high endurance, low power consumption, low latency, and CMOS compatibility—are a promising technology for memory, and domain-wall magnetic tunnel junction (DW-MTJ) devices have been shown to implement synaptic functions such as long-term potentiation and spike-timing dependent plasticity. In this work, we propose a notched DW-MTJ synapse as a candidate for supervised learning. Using micromagnetic simulations at room temperature, we show that notched synapses ensure the non-volatility of the synaptic weight and allow for highly linear, symmetric, and reproducible weight updates using either spin transfer torque (STT) or spin–orbit torque (SOT) mechanisms of DW propagation. We use lookup tables constructed from micromagnetics simulations to model the training of neural networks built with DW-MTJ synapses on both the MNIST and Fashion-MNIST image classification tasks. Accounting for thermal noise and realistic process variations, the DW-MTJ devices achieve classification accuracy close to ideal floating-point updates using both STT and SOT devices at room temperature and at 400 K. Our work establishes the basis for a magnetic artificial synapse that can eventually lead to hardware neural networks with fully spintronic matrix operations implementing machine learning.

42 ENGINEERING↗
DOE OSTI2024

HDBind: encoding of molecular structure with hyperdimensional binary representations

Traditional methods for identifying “hit” molecules from a large collection of potential drug-like candidates rely on biophysical theory to compute approximations to the Gibbs free energy of the binding interaction between the drug and its protein target. These approaches have a significant limitation in that they require exceptional computing capabilities for even relatively small collections of molecules. Increasingly large and complex state-of-the-art deep learning approaches have gained popularity with the promise to improve the productivity of drug design, notorious for its numerous failures. However, as deep learning models increase in their size and complexity, their acceleration at the hardware level becomes more challenging. Hyperdimensional Computing (HDC) has recently gained attention in the computer hardware community due to its algorithmic simplicity relative to deep learning approaches. The HDC learning paradigm, which represents data with high-dimension binary vectors, allows the use of low-precision binary vector arithmetic to create models of the data that can be learned without the need for the gradient-based optimization required in many conventional machine learning and deep learning methods. This algorithmic simplicity allows for acceleration in hardware that has been previously demonstrated in a range of application areas (computer vision, bioinformatics, mass spectrometery, remote sensing, edge devices, etc.). To the best of our knowledge, our work is the first to consider HDC for the task of fast and efficient screening of modern drug-like compound libraries. We also propose the first HDC graph-based encoding methods for molecular data, demonstrating consistent and substantial improvement over previous work. We compare our approaches to alternative approaches on the well-studied MoleculeNet dataset and the recently proposed LIT-PCBA dataset derived from high quality PubChem assays. We demonstrate our methods on multiple target hardware platforms, including Graphics Processing Units (GPUs) and Field Programmable Gate Arrays (FPGAs), showing at least an order of magnitude improvement in energy efficiency versus even our smallest neural network baseline model with a single hidden layer. Our work thus motivates further investigation into molecular representation learning to develop ultra-efficient pre-screening tools. We make our code publicly available at https://github.com/LLNL/hdbind.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

Neuromorphic intermediate representation: A unified instruction set for interoperable brain-inspired computing

Abstract Spiking neural networks and neuromorphic hardware platforms that simulate neuronal dynamics are getting wide attention and are being applied to many relevant problems using Machine Learning. Despite a well-established mathematical foundation for neural dynamics, there exists numerous software and hardware solutions and stacks whose variability makes it difficult to reproduce findings. Here, we establish a common reference frame for computations in digital neuromorphic systems, titled Neuromorphic Intermediate Representation (NIR). NIR defines a set of computational and composable model primitives as hybrid systems combining continuous-time dynamics and discrete events. By abstracting away assumptions around discretization and hardware constraints, NIR faithfully captures the computational model, while bridging differences between the evaluated implementation and the underlying mathematical formalism. NIR supports an unprecedented number of neuromorphic systems, which we demonstrate by reproducing three spiking neural network models of different complexity across 7 neuromorphic simulators and 4 digital hardware platforms. NIR decouples the development of neuromorphic hardware and software, enabling interoperability between platforms and improving accessibility to multiple neuromorphic technologies. We believe that NIR is a key next step in brain-inspired hardware-software co-evolution, enabling research towards the implementation of energy efficient computational principles of nervous systems. NIR is available atneuroir.org

Science & Technology - Other Topics↗
NASA NTRSDate not supplied

Initial Development of A Digital Twin Model for an Electrified Aircraft Propulsion Emulation Rig

In support of aviation fuel burn and emission reduction goals, NASA is pursing high-payoff research investments that promise to transform aviation. This includes investments in Electrified Aircraft Propulsion (EAP), which relies on the generation, storage, transmission, and use of electrical power for producing thrust and optimizing propulsion system efficiency. Multiple technology challenges must be addressed to unlock the full potential of EAP. This includes advances in propulsion controls, which will be vital for ensuring coordinated efficient operation of the complex integrated subsystems that comprise EAP architectures. To support EAP controls research, the NASA Glenn Research Center has developed the Hybrid Propulsion Emulation Rig (HyPER). The HyPER laboratory hardware includes shaft-mounted electric machines, power converters, power supplies, power distribution cables, and an energy storage device that can be reconfigured to represent a variety of EAP architectures. It also includes an integrated real-time computer system that hosts developed EAP control software and turbomachinery simulations. This enables the electrical system and rotating shafts of EAP designs to be implemented in actual hardware and integrated with turbomachinery simulations and system-level EAP control logic implemented in software. In this form, the HyPER laboratory provides a partially simulated, partially hardware-in-the-loop test environment enabling the initial development and evaluation of EAP control technology. A prerequisite for the development of EAP control designs is the availability of a system model that accurately reflects the operation of the electrical system hardware. To support this need, a digital twin model of the HyPER electrical system hardware is under development. This model is being coded in the MATLAB Simulink environment and uses the NASA-developed Electrical Modeling and Thermal Analysis Toolbox (EMTAT) to construct a digital twin framework. EMTAT contains generic electrical component building blocks that are simulated at turbomachinery timescales. Associated inputs and outputs allow the blocks to be combined to model complete electrical systems. The EMTAT blocks also contain adjustable internal maps and parameters that can be set to reflect the operation of a specific electrical component. For the HyPER digital twin, the settings of these EMTAT block internal maps and parameters is determined through machine learning approaches applied to characterization run data collected from the laboratory. During characterization runs the laboratory electrical system hardware is subjected to a full range of torque, speed, and power settings. Acquired data is then used to estimate EMTAT block parameters using a variety of machine learning techniques. The resulting digital twin model is found to match the operation of actual HyPER hardware with an accuracy suitable for control development purposes. It also holds promise for other applications including modeling the performance of HyPER laboratory reconfigurations and model-based anomaly detection. Planned follow-on work to automate post-processing of acquired laboratory data to update the HyPER digital twin model will also be presented and discussed.

Electrified Aircraft Propulsion↗
DOE OSTI2019

Multi-Objective Optimization for Size and Resilience of Spiking Neural Networks

Inspired by the connectivity mechanisms in the brain, neuromorphic computing architectures model Spiking Neural Networks (SNNs) in silicon. As such, neuromorphic architectures are designed and developed with the goal of having small, low power chips that can perform control and machine learning tasks. However, the power consumption of the developed hardware can greatly depend on the size of the network that is being evaluated on the chip. Furthermore, the accuracy of a trained SNN that is evaluated on chip can change due to voltage and current variations in the hardware that perturb the learned weights of the network. While efforts are made on the hardware side to minimize those perturbations, a software based strategy to make the deployed networks more resilient can help further alleviate that issue. In this work, we study Spiking Neural Networks in two neuromorphic architecture implementations with the goal of decreasing their size, while at the same time increasing their resiliency to hardware faults. We leverage an evolutionary algorithm to train the SNNs and propose a multiobjective fitness function to optimize the size and resiliency of the SNN. We demonstrate that this strategy leads to well-performing, small-sized networks that are more resilient to hardware faults.

Dimovska, Mihaela↗
DOE OSTI2023

PyTorch Circuit-Aware Bit-Cell Modeling

SAND2024-08549O PyTorch Circuit-Aware Bit-Cell Modeling, a Python library, demonstrates circuit-aware training for analog machine learning accelerators. Built upon PyTorch, this software provides hardware-aware versions of common neural network layers to mitigate non-ideal behavior from novel hardware. The software provides accurate forward and backward pass estimates for hardware-mapped neural networks that are implemented in crossbars (arrays) that contain a non-linear transistor selector device. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.

Bennett, Christopher↗
NASA NTRS1994

(abstract) Student Involvement in the Pluto Mission

The Pluto Fast Flyby mission development baseline consists of 2 identical spacecraft (120 - 165 kg) to be launched to Pluto/ Charon in the late 1990s. These spacecraft are intended to fly by Pluto and Charon in order to perform various remote-sensing scientific investigations and have a mission development cost less than $400M (FY92$) through launch plus 30 days. The Pluto team is committed to involving students in all areas of mission development and operations. In November 1992, the Pluto team sent a request for information to industry and universities looking for ways to lower the mass and cost of the mission. A number of universities responded with creative and promising technological developments. In addition to contracts with industry and other federal labs, contracts were signed with schools which allowed students to apply their research, enabling the Pluto team to use valuable resources on a variety of advanced technology endeavors. Perhaps the most exciting aspect of these investigations was that the deliverables that the students produced were not just final reports, but actual prototype hardware complete with write-ups on lessons learned in machining, programming, and design. Another exciting development was a prototype adapter competition in which 7 universities competed to design, build, and test their idea of a lightweight spacecraft-propulsion stack adapter. Georgia Tech won with an innovative dodecahedron composite lattice cone. Other students from other universities were involved as well. All in all, over 40 students from 20 different colleges made significant contributions to the Pluto Fast Flyby mission development through their efforts. This paper will give an overview of Pluto student involvement, the technologies which they examined, and useful results for the mission.

Pluto student involvement design universities tech↗
NASA NTRS2013

Personnel Data Analysis and Retrieval of Phase 1 Move To LC-39 Area

As a technology major from Jackson State University (JSU) I was called in as a summer intern at Kennedy Space Center (KSC) to work in the NASA Engineering, Control and Data Systems (NE-C) Division supporting the Spaceport Command and Control System (SCCS) at the Space Station Processing Facility (SSPF). I was given a two-part project; the first consisted of lending support relocating SCCS Computer Equipment and Project Personnel to the Launch Control Center (LCC). This task involved me using a Microsoft Office data processing tool to assist with the analysis and information management of logistics worth millions of dollars. With the assistance of two other interns, I was responsible for collecting data on equipment used, on a daily basis, by over 200 KSC employees. The many network servers, enterprise switches, desktop computers, and fiber optics had to be handled in an equally prompt and precise manner in order to ensure a minimal amount of equipment down time; which is critical in ensuring a properly secured networking environment. The second part of my project was to assist KSC in developing a more cost effective way of maintaining and taking full advantage of the functionality of some new kiosk units. Since KSC currently has no expert on the servicing and maintenance of the units, I, as a computer technology major, was given the opportunity to assess the hardware and software of the machines. The goal was to learn to establish a secure and remote environment for the kiosks; a goal highly valuing convenience by preserving valuable man-hours saved by not having to travel to each individual kiosk location. In addition, I was to leave a clear and precise plan for future users and administrators of the devices to follow.

Davis, Derrick D.↗
NASA NTRS2014

Research Data Alliance: Understanding Big Data Analytics Applications in Earth Science

The Research Data Alliance (RDA) enables data to be shared across barriers through focused working groups and interest groups, formed of experts from around the world - from academia, industry and government. Its Big Data Analytics (BDA) interest groups seeks to develop community based recommendations on feasible data analytics approaches to address scientific community needs of utilizing large quantities of data. BDA seeks to analyze different scientific domain applications (e.g. earth science use cases) and their potential use of various big data analytics techniques. These techniques reach from hardware deployment models up to various different algorithms (e.g. machine learning algorithms such as support vector machines for classification). A systematic classification of feasible combinations of analysis algorithms, analytical tools, data and resource characteristics and scientific queries will be covered in these recommendations. This contribution will outline initial parts of such a classification and recommendations in the specific context of the field of Earth Sciences. Given lessons learned and experiences are based on a survey of use cases and also providing insights in a few use cases in detail.

Riedel, Morris↗
DOE OSTI2023

Machine learning reduces soft costs for residential solar photovoltaics

Further deployment of rooftop solar photovoltaics (PV) hinges on the reduction of soft (non-hardware) costs—now larger and more resistant to reductions than hardware costs. The largest portion of these soft costs is the expenses solar companies incur to acquire new customers. In this study, we demonstrate the value of a shift from significance-based methodologies to prediction-oriented models to better identify PV adopters and reduce soft costs. We employ machine learning to predict PV adopters and non-adopters, and compare its prediction performance with logistic regression, the dominant significance-based method in technology adoption studies. Our results show that machine learning substantially enhances adoption prediction performance: The true positive rate of predicting adopters increased from 66 to 87%, and the true negative rate of predicting non-adopters increased from 75 to 88%. We attribute the enhanced performance to complex variable interactions and nonlinear effects incorporated by machine learning. With more accurate predictions, machine learning is able to reduce customer acquisition costs by 15% ($0.07/Watt) and identify new market opportunities for solar companies to expand and diversify their customer bases. Our research methods and findings provide broader implications for the adoption of similar clean energy technologies and related policy challenges such as market growth and energy inequality.

14 SOLAR ENERGY↗
DOE OSTI2022

The SODA Approach: Leveraging High-Level Synthesis for Hardware/Software Co-design and Hardware Specialization: Invited

Novel "converged" applications combine phases of scientific simulation with data analysis and machine learning. Each computational phase can benefit from specialized accelerators. However, algorithms evolve so quickly that mapping them on existing accelerators is suboptimal or even impossible. This paper presents the SODA (Software Defined Accelerators) framework, a modular, multi-level, open-source, no-human-in-the-loop, hardware synthesizer that enables end-to-end generation of specialized accelerators. SODA is composed of SODA-Opt, a high-level frontend developed in MLIR that interfaces with domain-specific programming frameworks and allows performing system level design, and Bambu, a state-of-the-art high-level synthesis engine that can target different device technologies. The framework implements design space exploration as compiler optimization passes. We show how the modular, yet tight, integration of the high-level optimizer and lower-level HLS tools enables the generation of accelerators optimized for the computational patterns of converged applications. We then discuss some of the research opportunities that such a framework allows, including system-level design, profile driven optimization, and supporting new optimization metrics.

Bohm Agostini, Nicolas↗
DOE OSTI2018

Deepmod: An Over-the-Air Trainable Machine Modem for Resilient PHY Layer Communications

Traditional physical layer protocols (e.g. WiFi, WiMax, etc.) are well established and are often optimal in a wide variety of channel conditions including heterogenous links and in tactical communications. Unfortunately, this same optimality encourages ubiquity in wireless communication technology and enhances the potential for catastrophic cyber or physical attacks due to prolific knowledge of underlying physical layers. Any truly resilient communications protocol must be capable of immediate redeployment to meet quality of service (QoS) demands in a wide variety of possible channel media. This work proposes an approach to communications that is contrary to much traditional approaches in that processing blocks are generated real-time and only relevant to the particular channel medium being used. Rather than creating man-made ubiquitous blocks of signal processing, we examine using processing that is immediately expendable once it has been used. This is achieved through software-defined radios, and deep modulation, where system blocks are replaced with machine learning graphs that can be trained, used, and then discarded as needed. Simulation and experimental hardware show how deep modulation can converge to viable communications links, using the same machine intelligence, in vastly different channels.

Anderson, Adam L↗