Search NASA⌕ Search

SEARCH · Search NASA

Results for “information architectures”

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 415 records · Page 23
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Development of Analysis Methods that Integrate Numeric and Textual Equipment Reliability Data

Within the Light Water Reactor Sustainability (LWRS) program, the Risk-Informed Systems Analysis (RISA) Pathway is performing collaborative research on the development and deployment of technologies designed to assist operating nuclear power plants (NPPs) to reduce operating costs improve plant reliability and availability. One of the RISA research areas is focusing on the development of methods and tools designed to optimize plant operations (e.g., maintenance/replacement schedules, optimal maintenance postures for plant structures, systems, and components [SSCs]) in a manner that is more cost effective than current approaches and makes better use of available SSC health data. The Risk-Informed Asset Management (RIAM) project targets this research area by creating a direct bridge between component equipment reliability (ER) data and system engineer decision making regarding maintenance activity scheduling and component aging management. In this respect, one challenge that NPP system engineers are facing is that the amount of ER data being continuously generated is not only extremely large in size, but it comes in different forms: textual (e.g., condition or maintenance reports) and numeric (e.g., generated by monitoring systems). All these data elements provide them with valuable insights and information regarding: 1) the discovery of anomalous behaviors or degradation trends, 2) the identification of the possible causes behind such behaviors/trends, and 3) the prediction of their direct consequences. However, several challenges have proved to be roadblocks to this process. While some of these challenges are technical in nature (i.e., data are often distributed over several physical servers/databases), others are conceptual in nature: data elements come in different formats (e.g., numeric or textual), and measured values have different scales (e.g., vibration spectra and oil temperature). The activities performed by the RIAM project during FY23 directly tackles the need to simultaneously integrate the analysis of ER data in all its forms, numeric and textual. Note that such task has never been performed before due to the complexity of the systems under consideration but, most importantly, because of the technical challenges behind the harmonization of ER data formats and the lack of adequate computational methods to analyze them. Our approach borrows ideas and concepts from the medical field where integration of several data sources is vital to assist medical practitioners to perform correct diagnosis and indicate optimal treatments. In our view a NPP asset is equivalent to a patient in a medical context. The main difference is the complexity of a human body is a magnitude more complex when compared to typical assets commonly present in NPPs (e.g., centrifugal pumps, or motor operated valves). This simplifies our first requirement when analyzing heterogenous ER data formats: to put data into “context”. Context is here intended as the additional piece of information that is needed by ER data analysis tools to understand what these data elements are referring to, i.e., which king of knowledge they are generating. In our context, this knowledge can be translated into models that capture the form and functional architecture of assets/systems, their dependencies, and how they interact. These models actually emulate the knowledge that that NPP system engineers possess about assets and systems; this is their key of success when analyzing ER data, their challenge is ability to handle large amount of data. Here, we employ model-based system engineering (MBSE) models of systems and assets to represent and capture their architecture and functional, i.e. cause-effect, relations. Then, ER data elements are processed by identifying first of all which elements of the developed MBSE elements they are referring to. For numeric ER data this task is fairly easy since it is possible to precisely pinpoint what MBSE elements the corresponding sensor are observing (e.g., bearing temperature of a centrifugal pump). Task is much harder for textual data since the information contained in issue or maintenance reports needs to “be understood” by a computational tool. Here we called this process as “knowledge extraction”. Once again, we borrow the experience in the medical field where methods to extract knowledge from textual data have been developed in the past decade. The missing element for us is the availability of a complete dictionary of NPP related concepts (in addition to the MBSE models presented earlier) that can put “text into context”. In FY23, such dictionary has been developed along with all the computational elements required for knowledge extraction. Lastly, once numeric and textual ER data elements have been processed and “understood”, then the last step is the discovery of possible cause-effect relations among them. This is performed by observing if a logical connection through the MBSE models exists, and if the

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

LCLS Big Data Handling – How I Learned to Stop Worrying and Love the Data Deluge

Advanced data and computing systems are vital to Linac Coherent Light Source (LCLS) operations, data interpretation and overall scientific productivity. The transition to MHz-era operation marks a fundamental change in scale that requires new infrastructure and architectures to link LCLS to the required scale of computing needed for scientific interpretation. The LCLS-II Data System meets big data challenges by implementing configurable data reduction that can adapt to multiple science areas, real-time analysis frameworks to provide visualization and fast feedback, and the ability to transfer data to local and remote computational facilities for near real time analysis at the appropriate scale. Feature extracted information generated in the data analysis pipeline - at the edge, local compute, or remote High-Performance Computing (HPC) resources - can be used to steer experiments and inform user decisions during beam time. Artificial Intelligence and Machine Learning (AI/ML) techniques present new opportunities to rapidly analyse large datasets and direct experiments, but create new challenges in scaling, adaptability, complexity, and trustworthiness. We describe how the LCLS-II Data System architecture addresses its data-driven challenges in the areas of data acquisition, data processing, data management, and workflow orchestration to decrease the overall time-to-science and provide a vision for future developments.

artificial intelligence↗
DOE OSTI2020

Steering Molecular Dynamics Simulations of Membrane-Associated Proteins with Neutron Reflection Results

Herein, we present a novel method to incorporate structural results from surface-sensitive scattering, such as X-ray or neutron reflectometry, into molecular dynamics simulations. While reflectometry techniques generally provide a means to determine the molecular-scale structures of organized interfacial films, they were recently shown to offer the capability to characterize the structures of protein-membrane complexes supported by a solid substrate. One-dimensional information inherent in the experimental results is used in the form of component volume occupancy (CVO) profiles, which describe the distribution of molecular components within an interfacial architecture, to construct real-space constraints in the form of a biasing potential for the simulation that vanishes when the simulated and experimental profiles agree. This approach improves the correspondence between simulation and experiment, as shown in the re-evaluation of an neutron-reflection-derived structure which was approximated by an independent molecular dynamics simulation in earlier work, and it also leads to faster equilibration of ensemble structures. We further show that time averaging the CVO profile that develops in the simulation while biasing with this approach permits fluctuations about the average that are necessary for conformational exploration of the system. This method is particularly valuable for studies of proteins at interfaces that contain disordered regions since the conformation of such regions is difficult to judge from the analysis of one-dimensional experimental profiles and may take prohibitively long to equilibrate in simulations.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Model Assumptions and Data Characteristics: Impacts on Domain Adaptation in Building Segmentation

Studies on domain adaptation (DA) for remote sensing (RS) imagery analysis lack consistency in selection and description of evaluation scenarios. Without properly characterizing datasets, model assumptions, and evaluation scenarios, it is difficult to objectively compare DA methods and reach conclusions about their suitability across different applications. With this motivation, this work seeks to empirically assess to which extent the interaction between data characteristics and model assumptions influences the effectiveness of DA methods. Using the widely explored task of building footprint segmentation as a case study, we perform a large-scale study across over 200 DA scenarios that include variations across view angles, areas observed, and sensors used for data acquisition. Rather than adopting different model architectures or optimization criteria, we contrast the performances of two DA methods based on adversarial learning that differ only in their assumptions about source and target domains. Informed by metadata and data characteristics unveiled using traditional computer vision (CV) techniques as well as pretrained deep models, we provide a detailed meta-analysis of experiments highlighting the importance of accurately considering data assumptions for DA in RS segmentation tasks. As demonstrated by a “cherry-picking” exercise, different claims regarding which model is best could be made by selecting different subsets of evaluation scenarios. While well-calibrated assumptions can be beneficial, mismatching assumptions can lead to negative biases in DA applications. Furthermore, this study intends to motivate the community toward more consistent evaluation protocols while providing recommendations and insights toward creating novel benchmark datasets, documenting data characteristics, application-specific knowledge, and model assumptions.

42 ENGINEERING↗
DOE OSTI2026

Oak Ridge National Laboratory Building Envelope Library (ORNOBEL)

The Oak Ridge National Laboratory Building Envelope Library (ORNOBEL) is a collection of dense exterior building-facade point clouds acquired using a survey-grade terrestrial laser scanner. Each file represents an individual facade from a building on the Oak Ridge National Laboratory (ORNL) campus or in Knoxville, Tennessee, with an average point-cloud resolution of approximately 3 mm. The points in each facade are semantically labeled into three classes: (1) window/door, representing openings in the building envelope; (2) wall, representing planar opaque envelope surfaces; and (3) other, representing the remaining facade-adjacent elements, architectural features, and protrusions. ORNOBEL supports the development, training, and evaluation of advanced deep-learning methods for automated building-envelope segmentation, geometric reconstruction, and building information modeling (BIM).

Maldonado Puente, Bryan [ORNL] (ORCID:000000033880↗
DOE OSTI2025

Track Reconstruction using Graph Neural Networks in the EMPHATIC Experiment

Track reconstruction is essential for extracting physics observables from detector data in high-energy and nuclear physics experiments. In this work, we investigate the use of graph neural networks (GNNs) to reconstruct particle momentum in the EMPHATIC experiment using simulated data. The model takes raw hit information from the silicon strip detectors (SSDs) as input and is trained to predict momentum components and the scattering angle of the particle. We describe the GNN architecture, training procedure, and performance metrics, and present results showing improved resolution in momentum reconstruction. These results demonstrate the potential of GNN-based approaches in track reconstruction tasks within complex detector environments like EMPHATIC.

Bhattarai, Aayush [Notre Dame U.]↗
DOE OSTI2021

Hardware-Based Emulator with Deep Learning Model for Building Energy Control and Prediction Based on Occupancy Sensors’ Data

Heating, ventilation, and air conditioning (HVAC) is the largest source of residential energy consumption. Occupancy sensors’ data can be used for HVAC control since it indicates the number of people in the building. HVAC and sensors form a typical cyber-physical system (CPS). In this paper, we aim to build a hardware-based emulation platform to study the occupancy data’s features, which can be further extracted by using machine learning models. In particular, we propose two hardware-based emulators to investigate the use of wired/wireless communication interfaces for occupancy sensor-based building CPS control, and the use of deep learning to predict the building energy consumption with the sensor data. We hypothesize is that the building energy consumption may be predicted by using the occupancy data collected by the sensors, and question what type of prediction model should be used to accurately predict the energy load. Another hypothesis is that an in-lab hardware/software platform could be built to emulate the occupancy sensing process. The machine learning algorithms can then be used to analyze the energy load based on the sensing data. To test the emulator, the occupancy data from the sensors is used to predict energy consumption. The synchronization scheme between sensors and the HVAC server will be discussed. We have built two hardware/software emulation platforms to investigate the sensor/HVAC integration strategies, and used an enhanced deep learning model—which has sequence-to-sequence long short-term memory (Seq2Seq LSTM)—with an attention model to predict the building energy consumption with the preservation of the intrinsic patterns. Because the long-range temporal dependencies are captured, the Seq2Seq models may provide a higher accuracy by using LSTM architectures with encoder and decoder. Meanwhile, LSTMs can capture the temporal and spatial patterns of time series data. The attention model can highlight the most relevant input information in the energy prediction by allocating the attention weights. The communication overhead between the sensors and the HVAC control server can also be alleviated via the attention mechanism, which can automatically ignore the irrelevant information and amplify the relevant information during CNN training. Our experiments and performance analysis show that, compared with the traditional LSTM neural network, the performance of the proposed method has a 30% higher prediction accuracy.

Ye, Zhijing↗
DOE OSTI2022

Augmented DNA Nanoarchitectures: A Structural Library of 3D Self-Assembling Tensegrity Triangle Variants

The DNA tensegrity triangle is known to reliably self-assemble into a 3D rhombohedral crystalline lattice via sticky-end cohesion. Here, in this study, the library of accessible motifs is expanded through covalent extensions of intertriangle regions and sticky-end-coordinated linkages of adjacent triangles with double helical segments using both geometrically symmetric and asymmetric configurations. The molecular structures of 18 self-assembled architectures at resolutions of 3.32–9.32 Å are reported; the observed cell dimensions, cavity sizes, and cross-sectional areas agree with theoretical expectations. These data demonstrate that fine control over triclinic and rhombohedral crystal parameters and the customizability of more complex 3D DNA lattices are attainable via rational design. It is anticipated that augmented DNA architectures may be fine-tuned for the self-assembly of designer nanocages, guest–host complexes, and proscriptive 3D nanomaterials, as originally envisioned. Finally, designer asymmetric crystalline building blocks can be seen as a first step toward controlling and encoding information in three dimensions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Continual Learning for Pattern Recognizers using Neurogenesis Deep Learning

Deep neural networks have emerged as a leading set of algorithms to infer information from a variety of data sources such as images and time series data. In their most basic form, neural networks lack the ability to adapt to new classes of information. Continual learning is a field of study attempting to give previously trained deep learning models the ability to adapt to a changing environment. Previous work developed a CL method called Neurogenesis for Deep Learning (NDL). Here, we combine NDL with a specific neural network architecture (the Ladder Network) to produce a system capable of automatically adapting a classification neural network to new classes of data. The NDL Ladder Network was evaluated against other leading CL methods. While the NDL and Ladder Network system did not match the cutting edge performance achieved by other CL methods, in most cases it performed comparably and is the only system evaluated that can learn new classes of information with no human intervention.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

A Cast of Thousands: How the IDEAS Productivity Project Has Advanced Software Productivity and Sustainability

Computational and data-enabled science and engineering are revolutionizing advances throughout science and society, at all scales of computing. For example, teams in the U.S. Department of Energy’s Exascale Computing Project have been tackling new frontiers in modeling, simulation, and analysis by exploiting unprecedented exascale computing capabilities—building an advanced software ecosystem that supports next-generation applications and addresses disruptive changes in computer architectures. However, concerns are growing about the productivity of the developers of scientific software. Members of the Interoperable Design of Extreme-scale Application Software project serve as catalysts to address these challenges through fostering software communities, incubating and curating methodologies and resources, and disseminating knowledge to advance developer productivity and software sustainability. This article discusses how these synergistic activities are advancing scientific discovery—mitigating technical risks by building a firmer foundation for reproducible, sustainable science at all scales of computing, from laptops to clusters to exascale and beyond.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

A Comparative Study of Physics‐Informed and Data‐Driven Neural Networks for Compound Flood Simulation at River‐Ocean Interfaces: A Case Study of Hurricane Irene

Simulating compound flooding (CF) at the river-ocean interface within large-scale Earth System Models (ESMs) presents significant challenges due to complex interactions between river discharge, storm surge, and tides. This study assesses the comparative advantages of physics-informed and data-driven machine learning (ML) approaches for enhancing local ESM performance. We systematically compare data-driven neural network models (i.e., CNNs, U-Net, Long Short-Term Memory (LSTM), Gated Recurrent Unit), and physics-informed neural network (PINN) models, including vanilla PINN and a finite-difference-based PINN (FD-PINN). Specifically, FD-PINN is introduced to enhance computational efficiency, accelerating vanilla PINNs by ∼6.5 times while improving accuracy. To enhance data-driven model training, a new data-generation approach is developed to sample historical fluvial and coastal flood events, which ensures a robust data set for extreme event prediction. The models are evaluated using a realistic one-dimensional river domain extracted from an ESM's river mesh and the Hurricane Irene event as an independent test case. Results show that FD-PINN achieves accurate predictions with significantly reduced computational costs relative to vanilla PINNs. Among data-driven models, the best overall performance is achieved by a CNN-LSTM hybrid, which balances accuracy and efficiency. While a fully connected CNN (CNN-FC) provides the best accuracy, it incurs high computational cost. Architectures lacking strong temporal modeling tend to underperform on unseen events. These findings highlight the importance of sequence-aware designs for robust generalization. This study reveals the trade-offs between physics-informed and data-driven models and proposes an adaptive hybrid framework for integrating ML into ESMs to enhance local flood simulations.

Earth Systems Modeling↗
DOE OSTI2025

Dynamical Signatures of Thermotoga maritima Maltose-Binding Proteins Affected by Ligand Binding

Functional segregation among protein isoforms depends on the interplay of their overall structures and the molecular dynamics of these structures. Thermotoga maritima maltose-binding protein (tmMBP) isoforms show size-dependent differential binding of maltose and malto-oligomers while maintaining remarkable fold conservation. This differential behavior needs detailed characterization in native-like aqueous conditions to understand the effects of protein dynamics on ligand binding and recognition. Small-angle neutron scattering (SANS), neutron spin echo (NSE) spectroscopy, and dynamic light scattering (DLS) were used in conjunction with previously published computational molecular dynamics (MD) simulations to understand the dynamic behavior of tmMBPs experimentally. SANS provided information on the overall structure of the molecules, while NSE was used to determine the dynamics in the nanosecond time scale. Both tmMBP2 and tmMBP3 have a bidomain architecture linked with a flexible hinge, with the binding pocket sitting in the cleft between the two domains. tmMBP2 and tmMBP3 showed different solution dynamics, with the translational and rotational components dominating the dynamics of both systems, resulting in a clear differentiation of their diffusion pattern. A faster dynamics component was also observed and was attributed to segmental dynamics. Differences observed between the ligand-free (apo) and ligand-bound (holo) states of the two proteins are attributed to conformational entropy. Our results highlight the intricacies of how structure and dynamics can together shape binding to a repertoire of substrates in structurally similar proteins.

Diffusion↗
DOE OSTI2024

A Privacy-Preserving Cyber Threat Intelligence Sharing System

Cyber Threat Intelligence (CTI) is a key resource for developing defensive strategies against potential cyber adversaries. Entities typically access CTI through open-source platforms, national agencies, or specialized commercial services. However, the bi-directional exchange of CTI is hindered by organizational trust boundaries, which complicate the sharing processes between entities and CTI providers. Centralized CTI services benefit from receiving suspicious cyber observables such as IP addresses, domain names, and email addresses from various entities. The aggregation allows for the correlation of widespread adversarial activities to enhance the alert and response mechanisms across the network of involved parties. Despite these benefits, openly sharing such observables incurs potential legal, regulatory, and reputational risks for the disclosing entities.This paper introduces a system designed to facilitate the secure exchange of cyber observables across trust boundaries without compromising the anonymity of the sharing entities. Here, we propose an architecture that leverages common web protocols alongside zero-knowledge proofs to authenticate members while maintaining anonymity. Additionally, we outline a privacy model tailored for STIX (Structured Threat Information eXpression) cyber observables to minimize the risk of inadvertently disclosing private information. Through our threat models, we assess the privacy implications of our proposed system and demonstrate its potential to enhance collaborative cyber defense efforts without exposing entities to undue risk.

BBS+ Signatures↗
DOE OSTI2022

A Phenome-Wide Association Study of genes associated with COVID-19 severity reveals shared genetics with complex diseases in the Million Veteran Program

The study aims to determine the shared genetic architecture between COVID-19 severity with existing medical conditions using electronic health record (EHR) data. We conducted a Phenome-Wide Association Study (PheWAS) of genetic variants associated with critical illness (n = 35) or hospitalization (n = 42) due to severe COVID-19 using genome-wide association summary data from the Host Genetics Initiative. PheWAS analysis was performed using genotype-phenotype data from the Veterans Affairs Million Veteran Program (MVP). Phenotypes were defined by International Classification of Diseases (ICD) codes mapped to clinically relevant groups using published PheWAS methods. Among 658,582 Veterans, variants associated with severe COVID-19 were tested for association across 1,559 phenotypes. Variants at the ABO locus (rs495828, rs505922) associated with the largest number of phenotypes (n rs495828 = 53 and n rs505922 = 59); strongest association with venous embolism, odds ratio (OR rs495828 1.33 (p = 1.32 x 10 –199 ), and thrombosis OR rs505922 1.33, p = 2.2 x10 -265 . Among 67 respiratory conditions tested, 11 had significant associations including MUC5B locus (rs35705950) with increased risk of idiopathic fibrosing alveolitis OR 2.83, p = 4.12 × 10 –191 ; CRHR1 (rs61667602) associated with reduced risk of pulmonary fibrosis, OR 0.84, p = 2.26× 10 –12 . The TYK2 locus (rs11085727) associated with reduced risk for autoimmune conditions, e.g., psoriasis OR 0.88, p = 6.48 x10 -23 , lupus OR 0.84, p = 3.97 x 10 –06 . PheWAS stratified by ancestry demonstrated differences in genotype-phenotype associations. LMNA (rs581342) associated with neutropenia OR 1.29 p = 4.1 x 10 –13 among Veterans of African and Hispanic ancestry but not European. Overall, we observed a shared genetic architecture between COVID-19 severity and conditions related to underlying risk factors for severe and poor COVID-19 outcomes. Differing associations between genotype-phenotype across ancestries may inform heterogenous outcomes observed with COVID-19. Divergent associations between risk for severe COVID-19 with autoimmune inflammatory conditions both respiratory and non-respiratory highlights the shared pathways and fine balance of immune host response and autoimmunity and caution required when considering treatment targets.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2025

Advanced Transmission Technologies – GETs and HPCs Session 1: ATT Foundations and Dynamic Line Ratings (DLRs)

The INL TADA GETs Cohort Session 1, held on November 4, 2025, convened experts to address the integration of advanced transmission technologies, including Grid-Enhancing Technologies (GETs) and High Performance Conductors (HPCs), with a focus on digital assurance challenges. The session highlighted the growing importance of cybersecurity, supply chain transparency, reliability, and business risk management in deploying GETs, especially Dynamic Line Ratings (DLRs). Participants examined how expanded attack surfaces, limited vendor pools, and new regulatory requirements—such as FERC Orders 881, 2023, and 1920—are influencing utilities and technology providers. The workshop underscored the need for cyber-informed engineering, secure-by-design principles, and practical risk management strategies, while fostering collaboration and knowledge sharing among industry peers. Technical discussions covered the evolution from static to dynamic line ratings, complexities of cloud-based architectures, and NERC CIP compliance challenges. The session concluded with a collaborative risk exercise and a preview of future workshops on advanced power flow control and transmission topology optimization, reinforcing the cohort’s commitment to advancing digital assurance in the energy sector.

24 - POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2024

Perspectives on AI Architectures and Codesign for Earth System Predictability

Abstract Recently, the U.S. Department of Energy (DOE), Office of Science, Biological and Environmental Research (BER), and Advanced Scientific Computing Research (ASCR) programs organized and held the Artificial Intelligence for Earth System Predictability (AI4ESP) workshop series. From this workshop, a critical conclusion that the DOE BER and ASCR community came to is the requirement to develop a new paradigm for Earth system predictability focused on enabling artificial intelligence (AI) across the field, laboratory, modeling, and analysis activities, called model experimentation (ModEx). BER’s ModEx is an iterative approach that enables process models to generate hypotheses. The developed hypotheses inform field and laboratory efforts to collect measurement and observation data, which are subsequently used to parameterize, drive, and test model (e.g., process based) predictions. A total of 17 technical sessions were held in this AI4ESP workshop series. This paper discusses the topic of the AI Architectures and Codesign session and associated outcomes. The AI Architectures and Codesign session included two invited talks, two plenary discussion panels, and three breakout rooms that covered specific topics, including 1) DOE high-performance computing (HPC) systems, 2) cloud HPC systems, and 3) edge computing and Internet of Things (IoT). We also provide forward-looking ideas and perspectives on potential research in this codesign area that can be achieved by synergies with the other 16 session topics. These ideas include topics such as 1) reimagining codesign, 2) data acquisition to distribution, 3) heterogeneous HPC solutions for integration of AI/ML and other data analytics like uncertainty quantification with Earth system modeling and simulation, and 4) AI-enabled sensor integration into Earth system measurements and observations. Such perspectives are a distinguishing aspect of this paper. Significance Statement This study aims to provide perspectives on AI architectures and codesign approaches for Earth system predictability. Such visionary perspectives are essential because AI-enabled model-data integration has shown promise in improving predictions associated with climate change, perturbations, and extreme events. Our forward-looking ideas guide what is next in codesign to enhance Earth system models, observations, and theory using state-of-the-art and futuristic computational infrastructure.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

Enabling Highly Efficient Capsule Networks Processing Through A PIM-Based Architecture Design

In recent years, the CNNs have achieved great successes in the image processing tasks, e.g., image recognition and object detection. Unfortunately, traditional CNN's classication is found to be easily misled by increasingly complex image features due to the usage of pooling operations, hence unable to preserve accurate position and pose information of the objects. To address this challenge, a novel neural network structure called Capsule Network has been proposed, which introduces equivariance through capsules to signicantly enhance the learning ability for image segmentation and object detection. Due to its requirement of performing a high volume of matrix operations, CapsNets have been generally accelerated on modern GPU platforms that provide highly optimized software library for common deep learning tasks. However, based on our performance characterization on modern GPUs, CapsNets exhibit low effciency due to the special program and execution features of their routing procedure, including massive unshareable intermediate variables and intensive syn- chronizations, which are very dicult to optimize at software level. To address these challenges, we propose a hybrid computing architecture design named PIM-CapsNet. It preserves GPU's on-chip computing capability for accelerating CNN types of layers in CapsNet, while pipelining with an off-chip in-memory acceleration solution that effectively tackles routing procedure's ineffciency by leveraging the processing-in-memory capability of today's 3D stacked memory. Using routing procedure's inherent parallellization feature, our design enables hierarchical improvements on CapsNet inference effciency through minimizing data movement and maximizing parallel processing in memory. Evaluation results demonstrate that our proposed design can achieve substantial improvement on both performance and energy savings for CapsNet inference, with almost zero accuracy loss. The results also suggest good performance scalability in optimizing the routing procedure with increasing network size.

Zhang, Xingyao↗
DOE OSTI2024

Perspectives on AI Architectures and Co-design for Earth System Predictability

Recently, the U.S. Department of Energy (DOE), Office of Science, Biological and Environmental Research (BER), and Advanced Scientific Computing Research (ASCR) programs organized and held the Artificial Intelligence for Earth System Predictability (AI4ESP) workshop series. From this workshop, a critical conclusion that the DOE BER and ASCR community came to is the requirement to develop a new paradigm for Earth system predictability focused on enabling artificial intelligence (AI) across the field, laboratory, modeling, and analysis activities, called model experimentation (ModEx). BER’s ModEx is an iterative approach that enables process models to generate hypotheses. The developed hypotheses inform field and laboratory efforts to collect measurement and observation data, which are subsequently used to parameterize, drive, and test model (e.g., process based) predictions. A total of 17 technical sessions were held in this AI4ESP workshop series. This paper discusses the topic of the AI Architectures and Codesign session and associated outcomes. The AI Architectures and Codesign session included two invited talks, two plenary discussion panels, and three breakout rooms that covered specific topics, including 1) DOE high-performance computing (HPC) systems, 2) cloud HPC systems, and 3) edge computing and Internet of Things (IoT). We also provide forward-looking ideas and perspectives on potential research in this codesign area that can be achieved by synergies with the other 16 session topics. These ideas include topics such as 1) reimagining codesign, 2) data acquisition to distribution, 3) heterogeneous HPC solutions for integration of AI/ML and other data analytics like uncertainty quantification with Earth system modeling and simulation, and 4) AI-enabled sensor integration into Earth system measurements and observations. Such perspectives are a distinguishing aspect of this paper.

58 GEOSCIENCES↗