Search NASA⌕ Search

SEARCH · Search NASA

Results for “multi-modal data”

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 73 records · Page 4
All sourcesNASA NTRSDOE OSTI
NASA NTRS2003

Enhanced Multi-Modal Access to Planetary Exploration

Tomorrow's Interplanetary Network (IPN) will evolve from JPL's Deep-Space Network (DSN) and provide key capabilities to future investigators, such as simplified acquisition of higher-quality science at remote sites and enriched access to these sites. These capabilities could also be used to foster public interest, e.g., by making it possible for students to explore these environments personally, eventually perhaps interacting with a virtual world whose models could be populated by data obtained continuously from the IPN. Our paper looks at JPL's approach to making this evolution happen, starting from improved communications. Evolving space protocols (e.g., today's CCSDS proximity and file-transfer protocols) will provide the underpinning of such communications in the next decades, just as today's rich web was enabled by progress in Internet Protocols starting from the early 1970's (ARPAnet research). A key architectural thrust of this effort is to deploy persistent infrastructure incrementally, using a layered service model, where later higher-layer capabilities (such as adaptive science planning) are enabled by earlier lower-layer services (such as automated routing of object-based messages). In practice, there is also a mind shift needed from an engineering culture raised on point-to-point single-function communications (command uplink, telemetry downlink), to one in which assets are only indirectly accessed, via well-defined interfaces. We are aiming to foster a 'community of access' both among space assets and the humans who control them. This enables appropriate (perhaps eventually optimized) sharing of services and resources to the greater benefit of all participants. We envision such usage to be as automated in the future as using a cell phone is today - with all the steps in creating the real-time link being automated.

interplanetary network (IPN)↗
DOE OSTI2026

Towards Secure Autonomous Vehicles: An Integrated Edge and Multi-Modal Machine Learning Framework for Intrusion Detection

Autonomous vehicles (AVs) are vulnerable to cyberattacks targeting both internal communication networks and external perception sensors. While edge-based intrusion de- tection for Controller Area Network (CAN) buses offers real-time protection, it cannot detect cross-modal threats. Conversely, multi-modal fusion approaches improve coverage but often lack efficiency for in-vehicle deployment. This thesis integrates two complemen- tary solutions: (1) a lightweight, edge-deployable machine learning framework for CAN bus intrusion detection, and (2) a late-fusion system combining CAN FD and LiDAR data. Together, they form a hierarchical defense capable of handling single-modality and coordi- nated attacks. Simulations show that CAN-only models reach 93% accuracy on simulated DoS, spoofing, replay, and fuzzy attacks, while the fusion system achieves 0.87 AUC and 0.82 F1-score at 2 ms latency. This unified framework establishes a scalable, explainable, and field-ready strategy for AV cybersecurity.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2016

Retrievals of Cloud Droplet Size from the RSP Data: Validation Using in Situ Measurements

We present comparisons of cloud droplet size distributions retrieved from the Research Scanning Polarimeter (RSP) data with correlative in situ measurements made during the North Atlantic Aerosols and Marine Ecosystems Study (NAAMES). This field experiment was based at St. Johns airport, Newfoundland, Canada with the latest deployment in May - June 2016. RSP was onboard the NASA C-130 aircraft together with an array of in situ and other remote sensing instrumentation. The RSP is an along-track scanner measuring polarized and total reflectances in9 spectral channels. Its unique high angular resolution allows for characterization of liquid water droplet size using the rainbow structure observed in the polarized reflectances in the scattering angle range between 135 and 165 degrees. A parametric fitting algorithm applied to the polarized reflectances provides retrievals of the droplet effective radius and variance assuming a prescribed size distribution shape (gamma distribution). In addition to this, we use a non-parametric method, Rainbow Fourier Transform (RFT), which allows us to retrieve the droplet size distribution (DSD) itself. The latter is important in the case of clouds with complex structure, which results in multi-modal DSDs. During NAAMES the aircraft performed a number of flight patterns specifically designed for comparison of remote sensing retrievals and in situ measurements. These patterns consisted of two flight segments above the same straight ground track. One of these segments was flown above clouds allowing for remote sensing measurements, while the other was at the cloud top where cloud droplets were sampled. We compare the DSDs retrieved from the RSP data with in situ measurements made by the Cloud Droplet Probe (CDP). The comparisons show generally good agreement with deviations explainable by the position of the aircraft within cloud and by presence of additional cloud layers in RSP view that do not contribute to the in situ DSDs. In the latter case the distributions retrieved from the RSP data were consistent with the multi-layer cloud structures observed in the correlative High Spectral Resolution Lidar (HSRL) profiles. The comparison results provide a rare validation of polarimetric droplet size retrieval techniques, which can be used for analysis of satellite data on global scale.

cloud droplets↗
DOE OSTI2026

LAF-Net: A Deep Residual and Cross-Attention Framework for Day-Ahead Load Forecasting: Preprint

Accurate day-ahead load forecasting is essential for reliable power system operations and market efficiency. System operators such as the Midcontinent Independent System Operator (MISO) rely on forecasts from multiple vendors, yet combining them effectively remains a persistent challenge due to vendor-specific biases. This paper presents a novel LSTM-Attention Fusion Network with Error Representation (LAF-Net) that enhances day-ahead hourly load forecasting through deep residual learning and multi-modal cross-attention. The proposed model builds a historical error memory from past vendor performance and dynamically queries it with future hour context to generate adaptive, hour-specific trust weights for each vendor. A bounded residual correction further refines forecasts by mitigating systematic and temporally localized errors. Tested on real MISO LBA data with multi-vendor forecasts, LAF-Net consistently outperforms the best vendor baseline across all 38 LBAs, achieving more than a 40% reduction in system-level mean absolute error (MAE) during peak load hours relative to the best vendor baseline.

24 POWER TRANSMISSION AND DISTRIBUTION↗
NASA NTRS2015

Integrated Assessments of the Impact of Climate Change on Agriculture: An Overview of AgMIP Regional Research in South Asia

South Asia encompasses a wide and highly varied geographic region, and includes climate zones ranging from the mountainous Himalayan territory to the tropical lowland and coastal zones along alluvial floodplains. The region's climate is dominated by a monsoonal circulation that heralds the arrival of seasonal rainfall, upon which much of the regional agriculture relies. The spatial and temporal distribution of this rainfall is, however, not uniform over the region. Northern South Asia, central India, and the west coast receive much of their rainfall during the southwest monsoon season, between June and September. These rains partly result from the moisture transport accompanying the monsoonal winds, which move in the southwesterly direction from the equatorial Indian Ocean. Regions further south, such as south/southeast India and Sri Lanka, may receive rains from both the southwest monsoon, and also during the northeast monsoon season between October and December (with northeasterly monsoon wind flow and moisture flux), which results in a bi- or multi-modal rainfall distribution. In addition, rainfall across South Asia displays a large amount of intraseasonal and interannual variability. Interannual variability is influenced by many drivers, both natural (e.g., El Ni-Southern Oscillation; ENSO) and man-made (e.g., rising temperatures due to increasing greenhouse gas concentrations), and it is challenging to obtaining accurate time-series of annual rainfall, even amongst various observed data products, which display inconsistencies amongst themselves. These climatic and rainfall variations can further complicate South Asia's agricultural and water management. Agriculture employs at least 65 of the workforce in most South Asian countries, and nearly 80 of South Asia's poor inhabit rural areas. Understanding the response of current agricultural production to climate variability and future climate change is of utmost importance in securing food and livelihoods for South Asia's growing population. In order to assess the future of food and livelihood security across South Asia, the Agricultural Model Intercomparison and Improvement Project (AgMIP) has undertaken integrated climate-crop-economic assessments of the impact of climate change on food security and poverty in South Asia, encompassing Bangladesh, India, Nepal, Pakistan, and Sri Lanka. AgMIP has funded, on a competitive basis, four South Asian regional research teams (RRTs) and one South Asian coordination team (CT) to undertake climate-crop-economic integrated assessments of food security for many districts in each of these countries, with the goal of characterizing the state of food security and poverty across the region, and projecting how these are subject to change under future climate change conditions.

mountains↗
NASA NTRS2015

Liquid Water Cloud Properties During the Polarimeter Definition Experiment (PODEX)

We present retrievals of water cloud properties from the measurements made by the Research Scanning Polarimeter (RSP) during the Polarimeter Definition Experiment (PODEX) held between January 14 and February 6, 2013. The RSP was onboard the high-altitude NASA ER-2 aircraft based at NASA Dryden Aircraft Operation Facility in Palmdale, California. The retrieved cloud characteristics include cloud optical thickness, effective radius and variance of cloud droplet size distribution derived using a parameter-fitting technique, as well as the complete droplet size distribution function obtained by means of Rainbow Fourier Transform. Multi-modal size distributions are decomposed into several modes and the respective effective radii and variances are computed. The methodology used to produce the retrieval dataset is illustrated on the examples of a marine stratocumulus deck off California coast and stratus/fog over California's Central Valley. In the latter case the observed bimodal droplet size distributions were attributed to two-layer cloud structure. All retrieval data are available online from NASA GISS website.

Electromagnetic scattering↗
DOE OSTI2025

Multi-modality deep learning for pulse prediction in homogeneous nonlinear systems via parametric conversion

In this Letter, we introduce FusionNet, a multi-modality deep learning framework designed to predict and analyze output pulses in high-power rare-earth-doped laser systems driving parametric conversion in homogeneous guided nonlinear media. FusionNet integrates temporal, spectral, and physical experimental conditions to model ultrafast nonlinear phenomena, including parametric nonlinear frequency conversion, self-phase modulation, and cross-phase modulation in homogeneous guided systems such as gas-filled hollow-core fibers. These systems bridge physical models with experimental data, advancing our understanding of light-guiding principles and nonlinear interactions while expediting the design and optimization of on-demand high-power, high-brightness systems. Our results demonstrate a 73% reduction in prediction error and an 83% improvement in computational efficiency compared to conventional neural networks. This work establishes a new paradigm for accelerating parametric simulations and optimizing experimental designs in high-power laser systems, with further implications for high-precision spectroscopy, quantum information science, and distributed entangled interconnects.

47 OTHER INSTRUMENTATION↗
DOE OSTI2026

Throughput Estimation of Data Transport Networks From Digital Twin Measurements

Digital twins of networked infrastructures, known as Virtual Infrastructure Twins (VITs), are increasingly used for software development, pre-deployment testing, and design space exploration. While VITs avoid the costs and potential disruptions associated with experiments on operational networks, their throughput measurements are typically not sufficiently accurate for performance profiling of wide-area networks that they emulate. Here, machine learning (ML) methods are developed to transform these inaccurate VIT network throughput measurements to closely match in peak and overall profile of those from a physical testbed or production network. First, a micro kernel network reflecting a physical network is utilized to collect one-time measurements on a host to support this ML transformation. Then, a generic multi-modal ML method is developed to learn a map that transforms measurements from subsequent VITs on the same host to match past, current and follow-on testbed and cloud networks. ML generalization equations are derived to establish its correctness and probabilistically guarantee its generalization accuracy. Experimental results are presented for a variety of VIT hosts with target testbed and cloud networks; they include a case study of a four-site science ecosystem wherein inaccurate convex VIT measurement profiles are transformed into accurate concave profiles of target networks.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS2005

Inductive Learning Approaches for Improving Pilot Awareness of Aircraft Faults

Neural network flight controllers are able to accommodate a variety of aircraft control surface faults without detectable degradation of aircraft handling qualities. Under some faults, however, the effective flight envelope is reduced; this can lead to unexpected behavior if a pilot performs an action that exceeds the remaining control authority of the damaged aircraft. The goal of our work is to increase the pilot s situational awareness by informing him of the type of damage and resulting reduction in flight envelope. Our methodology integrates two inductive learning systems with novel visualization techniques. One learning system, the Inductive Monitoring System (IMS), learns to detect when a simulation includes faulty controls, while two others, Inductive Classification System (INCLASS) and multiple binary decision tree system (utilizing C4.5), determine the type of fault. In off-line training using only non-failure data, IMS constructs a characterization of nominal flight control performance based on control signals issued by the neural net flight controller. This characterization can be used to determine the degree of control augmentation required in the pitch, roll, and yaw command channels to counteract control surface failures. This derived information is typically sufficient to distinguish between the various control surface failures and is used to train both INCLASS and C4.5. Using data from failed control surface flight simulations, INCLASS and C4.5 independently discover and amplify features in IMS results that can be used to differentiate each distinct control surface failure situation. In real-time flight simulations, distinguishing features learned during training are used to classify control surface failures. Knowledge about the type of failure can be used by an additional automated system to alter its approach for planning tactical and strategic maneuvers. The knowledge can also be used directly to increase the pilot s situational awareness and inform manual maneuver decisions. Our multi-modal display of this information provides speech output to issue control surface failure warnings to a lesser-used communication channel and provides graphical displays with pilot-selectable !eve!s of details to issues additional information about the failure. We also describe a potential presentation for flight envelope reduction that can be viewed separately or integrated with an existing attitude indicator instrument. Preliminary results suggest that the inductive approach is capable of detecting that a control surface has failed and determining the type of fault. Furthermore, preliminary evaluations suggest that the interface discloses a concise summary of this information to the pilot.

Spikovska, Lilly↗
NASA NTRS2003

Performance Evaluation Modeling of Network Sensors

Substantial benefits are promised by operating many spatially separated sensors collectively. Such systems are envisioned to consist of sensor nodes that are connected by a communications network. A simulation tool is being developed to evaluate the performance of networked sensor systems, incorporating such metrics as target detection probabilities, false alarms rates, and classification confusion probabilities. The tool will be used to determine configuration impacts associated with such aspects as spatial laydown, and mixture of different types of sensors (acoustic, seismic, imaging, magnetic, RF, etc.), and fusion architecture. The QualNet discrete-event simulation environment serves as the underlying basis for model development and execution. This platform is recognized for its capabilities in efficiently simulating networking among mobile entities that communicate via wireless media. We are extending QualNet's communications modeling constructs to capture the sensing aspects of multi-target sensing (analogous to multiple access communications), unimodal multi-sensing (broadcast), and multi-modal sensing (multiple channels and correlated transmissions). Methods are also being developed for modeling the sensor signal sources (transmitters), signal propagation through the media, and sensors (receivers) that are consistent with the discrete event paradigm needed for performance determination of sensor network systems. This work is supported under the Microsensors Technical Area of the Army Research Laboratory (ARL) Advanced Sensors Collaborative Technology Alliance.

probability of detection↗
NASA NTRS2017

RNAV STAR Procedural Adherence

In this exploratory archival study we mined the performance of 24 major US airports area navigation standard terminal arrival routes (RNAV STARs) over the preceding three years. Overlaying radar track data on top of RNAV STAR routes provided a comparison between aircraft flight paths and the waypoint positions and altitude restrictions. NASA Ames Supercomputing resources were utilized to perform the data mining and processing. We investigated STARs by lateral transition path (full-lateral), vertical restrictions (full-lateral/full-vertical), and skipped waypoints (skips). In addition, we graphed altitudes and their frequencies of occurrence for altitude restrictions. Full-lateral compliance was generally greater than Full-lateral/full-vertical, but the delta between the rates was not always consistent. Full-lateral/full-vertical usage medians of the 2016 procedures ranged from 0 in KDEN (Denver) to 21 in KMEM (Memphis). Waypoint skips ranged from 0 to nearly 100 for specific waypoints. Altitudes restrictions were sometimes missed by systemic amounts in 1000 ft. increments from the restriction, creating multi-modal distributions. Other times, altitude misses looked to be more normally distributed around the restriction. This work is a preliminary investigation into the objective performance of instrument procedures and provides a framework to track how procedural concepts and design intervention function. In addition, this tool may aid in providing acceptability metrics as well as risk assessment information.

Stewart, Michael J.↗
DOE OSTI2025

SANE: strategic autonomous non-smooth exploration for multiple optima discovery in multi-modal and non-differentiable black-box functions

Both computational and experimental material discovery bring forth the challenge of exploring multidimensional and multimodal parameter spaces, such as phase diagrams of Hamiltonians with multiple interactions, composition spaces of combinatorial libraries, material structure image spaces, and molecular embedding spaces. Often these systems are black-boxes and time-consuming to evaluate, which resulted in strong interest towards active learning methods such as Bayesian optimization (BO). However, these systems are often noisy which make the black box function severely multi-modal and non-differentiable, where a vanilla BO can get overly focused near a single or faux optimum, deviating from the broader goal of scientific discovery. To address these limitations, here we developed Strategic Autonomous Non-Smooth Exploration (SANE) to facilitate an intelligent Bayesian optimized navigation with a proposed cost-driven probabilistic acquisition function to find multiple global and local optimal regions, avoiding the tendency to becoming trapped in a single optimum. To distinguish between a true and false optimal region due to noisy experimental measurements, a human (domain) knowledge driven dynamic surrogate gate is integrated with SANE. We implemented the gate-SANE into pre-acquired piezoresponse spectroscopy data of a ferroelectric combinatorial library with high noise levels in specific regions, and piezoresponse force microscopy (PFM) hyperspectral data. SANE demonstrated better performance than classical BO to facilitate the exploration of multiple optimal regions and thereby prioritized learning with higher coverage of scientific values in autonomous experiments. Our work showcases the potential application of this method to real-world experiments, where such combined strategic and human intervening approaches can be critical to unlocking new discoveries in autonomous research.

Biswas, Arpan [University of Tennessee, Knoxville,↗
NASA NTRS2017

Objective Assessment Method for RNAV STAR Adherence

Flight crews and air traffic controllers have reported many safety concerns regarding area navigation standard terminal arrival routes (RNAV STARs). Specifically, optimized profile descents (OPDs). However, our information sources to quantify these issues are limited to subjective reporting and time consuming case-by-case investigations. This work is a preliminary study into the objective performance of instrument procedures and provides a framework to track procedural concepts and assess design specifications. We created a tool and analysis methods for gauging aircraft adherence as it relates to RNAV STARs. This information is vital for comprehensive understanding of how our air traffic behaves. In this study, we mined the performance of 24 major US airports over the preceding three years. Overlaying 4D radar track data onto RNAV STAR routes provided a comparison between aircraft flight paths and the waypoint positions and altitude restrictions. NASA Ames Supercomputing resources were utilized to perform the data mining and processing. We assessed STARs by lateral transition path (full-lateral), vertical restrictions (full-lateral/full-vertical), and skipped waypoints (skips). In addition, we graphed frequencies of aircraft altitudes relative to the altitude restrictions. Full-lateral adherence was always greater than Full-lateral/ full- vertical, as it is a subset, but the difference between the rates was not consistent. Full-lateral/full-vertical adherence medians of the 2016 procedures ranged from 0% in KDEN (Denver) to 21% in KMEM (Memphis). Waypoint skips ranged from 0% to nearly 100% for specific waypoints. Altitudes restrictions were sometimes missed by systematic amounts in 1,000 ft. increments from the restriction, creating multi-modal distributions. Other times, altitude misses looked to be more normally distributed around the restriction. This tool may aid in providing acceptability metrics as well as risk assessment information.

Stewart, Michael↗
NASA NTRS2003

Development of a Dynamically Configurable, Object-Oriented Framework for Distributed, Multi-modal Computational Aerospace Systems Simulation: Second Year Progress Report

Mesh generation has long been recognized as a bottleneck in the CFD process. While much research on automating the volume mesh generation process have been relatively successful,these methods rely on appropriate initial surface triangulation to work properly. Surface discretization has been one of the least automated steps in computational simulation due to its dependence on implicitly defined CAD surfaces and curves. Differences in CAD peometry engines manifest themselves in discrepancies in their interpretation of the same entities. This lack of "good" geometry causes significant problems for mesh generators, requiring users to "repair" the CAD geometry before mesh generation. The problem is exacerbated when CAD geometry is translated to other forms (e.g., IGES )which do not include important topological and construction information in addition to entity geometry. One technique to avoid these problems is to access the CAD geometry directly from the mesh generating software, rather than through files. By accessing the geometry model (not a discretized version) in its native environment, t h s a proach avoids translation to a format which can deplete the model of topological information. Our approach to enable models developed in the Denali software environment to directly access CAD geometry and functions is through an Application Programming Interface (API) known as CAPRI. CAPRI provides a layer of indirection through which CAD-specific data may be accessed by an application program using CAD-system neutral C and FORTRAN language function calls. CAPRI supports a general set of CAD operations such as truth testing, geometry construction and entity queries.

Afjeh, Abdollah A.↗
NASA NTRS2017

A 3D Active Learning Application for NeMO-Net, the NASA Neural Multi-Modal Observation and Training Network for Global Coral Reef Assessment

NeMO-Net, the NASA neural multi-modal observation and training network for global coral reef assessment, is an open-source deep convolutional neural network and interactive active learning training software aiming to accurately assess the present and past dynamics of coral reef ecosystems through determination of percent living cover and morphology as well as mapping of spatial distribution. We present an interactive video game prototype for tablet and mobile devices where users interactively label morphology classifications over mm-scale 3D coral reef imagery captured using fluid lensing to create a dataset that will be used to train NeMO-Nets convolutional neural network. The application currently allows for users to classify preselected regions of coral in the Pacific and will be expanded to include additional regions captured using our NASA FluidCam instrument, presently the highest-resolution remote sensing benthic imaging technology capable of removing ocean wave distortion, as well as lower-resolution airborne remote sensing data from the ongoing NASA CORAL campaign. Active learning applications present a novel methodology for efficiently training large-scale Neural Networks wherein variances in identification can be rapidly mitigated against control data. NeMO-Net periodically checks users input against pre-classified coral imagery to gauge their accuracy and utilize in-game mechanics to provide classification training. Users actively communicate with a server and are requested to classify areas of coral for which other users had conflicting classifications and contribute their input to a larger database for ranking. In partnering with Mission Blue and IUCN, NeMO-Net leverages an international consortium of subject matter experts to classify areas of confusion identified by NeMO-Net and generate additional labels crucial for identifying decision boundary locations in coral reef assessment.

NeMO=NET↗
DOE OSTI2026

Quantum-Inspired Bayesian Sampling for Uncertainty Quantification and Machine Learning (Final Technical Report)

With increasing simulation and measurement data, machine learning and artificial intelligence have been widely used in computational decision-making of complex engineering systems. The resulting tools, such as uncertainty quantification solvers, reinforcement learning, and physics-informed machine learning, have achieved great success in critical DOE tasks such as material discovery and design, energy system modeling and control, and numerical weather and climate prediction. A core topic in scientific machine learning and artificial intelligence is Bayesian inference: given an observed data set, people want to estimate the posterior distribution of a (possibly large) number of hidden parameters. Due to the flexibility and weak assumptions, Bayesian sampling has been the mainstream Bayesian inference solvers despite the rapid progress of approximate Bayesian inference. Classical Bayesian sampling methods such as Markov-chain Monte Carlo suffer from a low-acceptance rate due to the random walk nature, therefore state-of-the-art techniques use Hamiltonian Monte Carlo and its variants to efficiently draw posterior samples in a high dimension. The key idea of Hamiltonian Monte Carlo and its variants is to simulate the Hamiltonian dynamics of a classical particle with a fixed mass, and their performance significantly degrades when the posterior distribution is highly spiky or has multiple modes. Leveraging the idea of quantum physics, this project has investigated new theory, algorithms and applications of Bayesian inference (especially Bayesian sampling). The main results include: (1) novel quantum-inspired Bayesian sampling methods that can lead to better accuracy for challenging multi-modal or spiky distributions, (2) more scalable machine learning framework leveraging tensor-compressed Bayesian inference, and (3) Bayesian and sampling approaches for verifying the robustness of continuous and binary neural networks.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Scalable Multi-Facility Workflows for Artificial Intelligence Applications in Climate Research

Earth observation satellites and earth system models are sources of vast, multi-modal datasets that are invaluable for advancing climate and environmental research. However, their scale and complexity pose significant challenges for processing and analysis. In this paper we discuss our experiences in developing and using a scientific research application using an automated multi-facility workflow that orchestrates data collection, preprocessing, artificial intelligence (AI) inferencing, and data movement across diverse computational resources, leveraging the Advanced Computing Ecosystem Testbed at the Oak Ridge Leadership Computing Facility (OLCF). We demonstrate that our workflow can be seamlessly integrated and orchestrated across research facilities managed by different federal agencies, thus allowing users to extract new scientific insights from climate datasets. The experimental results indicate that the multi-facility workflow significantly reduces processing time, enhances scalability, and maintains high efficiency across varying workloads. Notably, our workflow processes 12,000 high-resolution satellite images in just 44 seconds using 80 workers distributed across 10 nodes on the OLCF systems. Such high throughput is essential for dynamic tokenization and sharding of petascale satellite data for distributed AI model training and inferencing at scale across thousands of GPUs.

Kurihana, Takuya [ORNL] (ORCID:0000000156698565)↗
NASA NTRS1989

A layered abduction model of perception: Integrating bottom-up and top-down processing in a multi-sense agent

A layered-abduction model of perception is presented which unifies bottom-up and top-down processing in a single logical and information-processing framework. The process of interpreting the input from each sense is broken down into discrete layers of interpretation, where at each layer a best explanation hypothesis is formed of the data presented by the layer or layers below, with the help of information available laterally and from above. The formation of this hypothesis is treated as a problem of abductive inference, similar to diagnosis and theory formation. Thus this model brings a knowledge-based problem-solving approach to the analysis of perception, treating perception as a kind of compiled cognition. The bottom-up passing of information from layer to layer defines channels of information flow, which separate and converge in a specific way for any specific sense modality. Multi-modal perception occurs where channels converge from more than one sense. This model has not yet been implemented, though it is based on systems which have been successful in medical and mechanical diagnosis and medical test interpretation.

Josephson, John R.↗