Results for “knowledge networks” — Page 29

DOE OSTI2025

Exploring Black-box Adversarial Attacks on Low-rank Constrained Neural Networks

Low-rank compression has been shown as an effective tool to reduce parameter counts of convolutional and vision transformer architectures; however, low-rank training often reduces model robustness to adversarial perturbations. In this work, we explore the effects of low-rank training on black-box attacks, where attacked images are generated without knowledge of the low-rank parameters. We find that low-rank training is not sufficient as a black-box defense and can sometimes produce worse than expected as compared to baseline models. Influencing the spectrum of the low-rank models during training, which is known to increase model robustness against white-box attacks, improves black-box performance as well.

Schnake, Stefan [ORNL] (ORCID:0000000215183538)↗

NASA NTRSDate not supplied

Exploring Ridesharing in Passenger Urban Air Mobility: A Comparative Analysis

There is growing interest in urban air mobility (UAM) as an alternative for passenger and cargo transport around metropolitan areas in a multimodal transportation system that leverages small, electric aircraft. Ridesharing has been proposed as a means of making UAM passenger trips more affordable and environmentally friendly. We present a UAM ridesharing model integrated into an existing computational framework for analyzing daily work commute trips within a metropolitan area. We leverage this model to estimate the potential demand for ridesharing-enabled UAM trips within six metropolitan areas across the United States: Chicago, IL; Cleveland, OH; Dallas, TX; Denver, CO; New York City, NY; and Orlando, FL. We compare results for each metropolitan area with and without ridesharing. Results indicate that ridesharing enables at least an order of magnitude more UAM-preferring passengers than without ridesharing, though specifics vary across metropolitan areas and network sizes. Enabling ridesharing in UAM also considerably lowers the mean and mode value of time for passengers that select the UAM mode, indicating that ridesharing can help make UAM more economically accessible to a larger set of the population. An important caveat is that the UAM ridesharing model does not account for operational constraints, such as aerodrome capacity and aircraft availability, and relies on a perfect knowledge of passenger movements and mode preferences. This leads to high UAM ridesharing volumes that are unlikely to reflect real-world UAM operations and thus serves as an upper bound estimate.

advanced air mobility↗

NASA NTRSDate not supplied

Exploring Ridesharing in Passenger Urban Air Mobility: A Comparative Analysis

There is growing interest in urban air mobility (UAM) as an alternative for passenger and cargo transport around metropolitan areas in a multimodal transportation system that leverages small, electric aircraft. Ridesharing has been proposed as a means of making UAM passenger trips more affordable and environmentally friendly. We present a UAM ridesharing model integrated into an existing computational framework for analyzing daily work commute trips within a metropolitan area. We leverage this model to estimate the potential demand for ridesharing-enabled UAM trips within six metropolitan areas across the United States: Chicago, IL; Cleveland, OH; Dallas, TX; Denver, CO; New York City, NY; and Orlando, FL. We compare results for each metropolitan area with and without ridesharing. Results indicate that ridesharing enables at least an order of magnitude more UAM-preferring passengers than without ridesharing, though specifics vary across metropolitan areas and network sizes. Enabling ridesharing in UAM also considerably lowers the mean and mode value of time for passengers that select the UAM mode, indicating that ridesharing can help make UAM more economically accessible to a larger set of the population. An important caveat is that the UAM ridesharing model does not account for operational constraints, such as aerodrome capacity and aircraft availability, and relies on a perfect knowledge of passenger movements and mode preferences. This leads to high UAM ridesharing volumes that are unlikely to reflect real-world UAM operations and thus serves as an upper bound estimate.

advanced air mobility↗

DOE OSTI2026

Development and Implementation of a New AI-Based Tool to Support Fast Reactor Software Model Generation and Validation

This report summarizes FY26 work to develop Maggie, an artificial intelligence-based assistant designed to support software model generation and validation activities for fast reactor analysis codes. The project established a modular, code-agnostic software architecture that separates reusable agent capabilities from code-specific knowledge and tools, with initial implementation focused on the FRP-supported fast reactor safety analysis code SAS4A/SASSYS1 (SAS). A curated SAS-specific knowledge base was assembled from the code manual, training materials, historical analysis reports, and representative input files, and was integrated through retrieval-augmented generation to ground Maggie’s responses in authoritative sources. Maggie was deployed on the internal Argonne network, where it demonstrated practical user-facing capability as a chatbot for answering natural language questions about SAS and retrieving relevant technical information. Demonstration cases also showed that Maggie can generate useful snippets of SAS input for selected modeling tasks, while highlighting current limitations in reliability and consistency for more complex input generation tasks. Overall, the FY26 effort established the technical foundation for an AI-assisted capability intended to improve the efficiency, consistency, and accessibility of fast reactor software model development at Argonne and, with further improvements, to support eventual use by the broader fast reactor community, including industry users of FRP-supported analysis tools.

Thomas, Rachel [Argonne National Laboratory (ANL),↗

NASA NTRS1991

An architecture for designing fuzzy logic controllers using neural networks

Described here is an architecture for designing fuzzy controllers through a hierarchical process of control rule acquisition and by using special classes of neural network learning techniques. A new method for learning to refine a fuzzy logic controller is introduced. A reinforcement learning technique is used in conjunction with a multi-layer neural network model of a fuzzy controller. The model learns by updating its prediction of the plant's behavior and is related to the Sutton's Temporal Difference (TD) method. The method proposed here has the advantage of using the control knowledge of an experienced operator and fine-tuning it through the process of learning. The approach is applied to a cart-pole balancing system.

Berenji, Hamid R.↗

NASA NTRS2007

The Determination of Titan Gravity Field from Doppler Tracking of the Cassini Spacecraft

In its tour of the Saturnian system, the spacecraft Cassini is carrying out measurements of the gravity field of Titan, whose knowledge is crucial for constraining the internal structure of the satellite. In the five flybys devoted to gravity science, the spacecraft is tracked in X (8.4 GHz) and Ka band (32.5 GHz) from the antennas of NASA's Deep Space Network. The use of a dual frequency downlink is used to mitigate the effects of interplanetary plasma, the largest noise source affecting Doppler measurements. Variations in the wet path delay are effectively compensated by means of advanced water vapor radiometers placed close to the ground antennas. The first three flybys occurred on February 27, 2006, December 28, 2006, and June 29, 2007. Two additional flybys are planned in July 2008 and May 2010. This paper presents the estimation of the mass and quadrupole field of Titan from the first two flybys, carried out by the Cassini Radio Science Team using a short arc orbit determination. The data from the two flybys are first independently fit using a dynamical model of the spacecraft and the bodies of the Saturnian system, and then combined in a multi-arc solution. Under the assumption that the higher degree harmonics are negligible, the estimated values of the gravity parameters from the combined, multi-arc solution are GM = 8978.1337 +/- 0.0025 km(exp 3) / s(exp 2), J (sub 2) = (2.7221 +/- 0.0185) 10 (exp -5) and C (sub 22) = (1.1159 +/- 0.0040) 10 (exp -5) The excellent agreement (within 1.7 sigma) of the results from the two flybys further increases the confidence in the solution and provides an a posteriori validation of the dynamical model.

Iess, L.↗

DOE OSTI2025

The power of lanthanides: same composition, but different lanthanides leading to different interesting materials properties, from magnetocalorics to molecular magnets and phosphors

Commonly accepted design concepts for ionic liquids (ILs) state that the constituting ions must be large and carry low, well-dispersed charges. A series of ILs based of pentadeca charged ILs with pentanuclear linear {Ln 5 } units ([Ln 5 (C 2 H 5 -C 3 H 3 N 2 -CH 2 COO) 16 (H 2 O) 8 ](Tf 2 N) 15 (C 3 H 3 N 2 = imidazolium moiety, Tf 2 N = bis(trifluoromethanesulfonyl)amide) with Ln = Er, Ho, Tm) demonstrates that these criteria are not absolute. Highly charged ions can also support IL formation, provided they are sufficiently large. Expanding the series of these unconventional, record pentadeca charged with new lanthanide representatives, led to the discovery of additional unprecedented properties for ILs: The Gd compound exhibits a strong magnetocaloric effect (MCE) in the liquid state with a maximum magnetic entropy change of −ΔS M = −11 J⋅kg −1 ⋅K −1 at 2 K for Δμ 0 H = 7 T. Albeit the Dy representative shows slow magnetic relaxation, the relaxation times are not favorable for practical application as a molecular magnet. Lastly, for both the Gd and the Y compound, phosphorescence in the seconds time scale is observed, which is, to the best of our knowledge, the longest ever reported for an IL.

Ionic Liquids↗

DOE OSTI2024

Chemical reaction enhanced graph learning for molecule representation

Abstract Motivation Molecular representation learning (MRL) models molecules with low-dimensional vectors to support biological and chemical applications. Current methods primarily rely on intrinsic molecular information to learn molecular representations, but they often overlook effectively integrating domain knowledge into MRL. Results In this article, we develop a reaction-enhanced graph learning (RXGL) framework for MRL, utilizing chemical reactions as domain knowledge. RXGL introduces dual graph learning modules to model molecule representation. One module employs graph convolutions on molecular graphs to capture molecule structures. The other module constructs a reaction-aware graph from chemical reactions and designs a novel graph attention network on this graph to integrate reaction-level relations into molecular modeling. To refine molecule representations, we design a reaction-based relation learning task, which considers the relations between the reactant and product sides in reactions. In addition, we introduce a cross-view contrastive task to strengthen the cooperative associations between molecular and reaction-aware graph learning. Experiment results show that our RXGL achieves strong performance in various downstream tasks, including product prediction, reaction classification, and molecular property prediction. Availability and implementation The code is publicly available at https://github.com/coder-ACAC/RLM.

Biochemistry & Molecular Biology↗

NASA NTRS1985

A discrete control model of PLANT

A model of the PLANT system using the discrete control modeling techniques developed by Miller is described. Discrete control models attempt to represent in a mathematical form how a human operator might decompose a complex system into simpler parts and how the control actions and system configuration are coordinated so that acceptable overall system performance is achieved. Basic questions include knowledge representation, information flow, and decision making in complex systems. The structure of the model is a general hierarchical/heterarchical scheme which structurally accounts for coordination and dynamic focus of attention. Mathematically, the discrete control model is defined in terms of a network of finite state systems. Specifically, the discrete control model accounts for how specific control actions are selected from information about the controlled system, the environment, and the context of the situation. The objective is to provide a plausible and empirically testable accounting and, if possible, explanation of control behavior.

Mitchell, C. M.↗

NASA NTRS1996

Sequence-of-events-driven automation of the deep space network

In February 1995, sequence-of-events (SOE)-driven automation technology was demonstrated for a Voyager telemetry downlink track at DSS 13. This demonstration entailed automated generation of an operations procedure (in the form of a temporal dependency network) from project SOE information using artificial intelligence planning technology and automated execution of the temporal dependency network using the link monitor and control operator assistant system. This article describes the overall approach to SOE-driven automation that was demonstrated, identifies gaps in SOE definitions and project profiles that hamper automation, and provides detailed measurements of the knowledge engineering effort required for automation.

Hill, R., Jr.↗

NASA NTRS1996

Sequence-of-Events-Driven Automation of the Deep Space Network

In February 1995, sequence-of-events (SOE)-driven automation technology was demonstrated for a Voyager telemetry downlink track at DSS 13. This demonstration entailed automated generation of an operations procedure (in the form of a temporal dependency network) from project SOE information using artificial intelligence planning technology and automated execution of the temporal dependency network using the link monitor and control operator assistant system. This article describes the overall approach to SOE-driven automation that was demonstrated, identifies gaps in SOE definitions and project profiles that hamper automation, and provides detailed measurements of the knowledge engineering effort required for automation.

Hill, R., Jr.↗

NASA NTRS2021

Collaborative Pose Estimation of An Unknown Target Using Multiple Spacecraft

A reliable method for pose estimation of an unknown and uncooperative space target using monocular vision remains an open problem. Vision-based pose determination can be challenging in case of unfavorable illumination, time-varying conditions due to rotational motion and relative orbit, and scale ambiguity resolution. To address these challenges, we propose a novel collaborative pose determination algorithm called Multi- Spacecraft Simultaneous Estimation of Pose and Shape algorithm or M-SEPS.Within M-SEPS, a team of chaser spacecraft, each equipped with a monocular camera, exchange information over a local network to jointly estimate the relative kinematic state of the target and its sparse shape landmarks. In this approach, each spacecraft processes its own images and observes particular target landmarks in parallel and in a distributed fashion. Then, the local network is exploited by the spacecraft to share their consensus proposals and aggregate them to achieve the joint estimate. We validate our algorithm using simulations of relative orbits and observations, captured by each chaser spacecraft. To the best of the authors’ knowledge, this is the first cooperative, vision-based algorithm for estimating the pose and shape of a space object for an arbitrary number of spacecraft.

Chung, Soon-Jo↗

DOE OSTI2025

Knowledge Oriented Graph Unified Transformer (KOGUT) v0.1

KOGUT — Knowledge Oriented Graph Unified Transformer KOGUT implements the Relational Graph Transformer (RelGT) architecture for knowledge graph link prediction in biological domains, with a primary focus on microbial growth media prediction. While the original RelGT (arXiv:2505.10960) targets relational tables, time series, and multi-table databases, KOGUT adapts this architecture for heterogeneous biological knowledge graphs, providing first-in-class AI predictive models for microbial cultivation. Key Adaptations Beyond Original RelGT: - Knowledge Graph Focus: Applied to biological KGs with semantic node types (taxa, chemicals, media, phenotypes, environments) versus generic relational database tables, trained on the KG-Microbe knowledge graph (1.3M entities, 2.9M edges, 24 relation types). - Multimodal Node Encoding: Integrates node labels, categories, descriptions, and synonyms from KG metadata through learned embedding layers—adapting relational column features to graph node attributes with textual semantics. - Extended K-Hop Subgraph Strategy: Optimized neighborhood sampling (3-hop default, configurable up to 200 nodes) tuned for sparse biological networks, building on the original local-global attention framework with biological relation preservation. - Biolink Predicate Preservation: Type-specific transformations for 24 biological edge semantics (occurs_in, consumes, produces, has_phenotype, subclass_of) beyond standard relational foreign keys, enabling multi-relation link prediction. - Inductive Learning Support: Enables zero-shot predictions for novel taxa through feature-based embeddings (temperature, oxygen requirements, gram stain, cell shape), extending the original transductive relational benchmark scope to uncultured microorganisms. CheapSOTA Performance Optimizations (This Distribution): - VQ-EMA Centroid Attention: Vector quantization with exponential moving average for improved global context modeling (+5-10% MRR improvement). - HDF5 Precomputed Data Loading: One-time preprocessing of k-hop subgraphs to eliminate redundant graph traversals (2-5× training speedup). - Distributed Data Parallel Training: Multi-GPU support for scaling to larger knowledge graphs (tested on 4× NVIDIA A100 GPUs at NERSC Perlmutter). - Mixed Precision Training: Automatic mixed precision (AMP) for memory efficiency and faster training. Advantages Over Standard Knowledge Graph Embedding Models: Combines RelGT's proven multi-element tokenization (features, type, hop, structure) with graph-native biological representations, enabling interpretable link prediction across heterogeneous entities that standard embedding models (TransE, RotatE, ComplEx) and table-based transformers cannot directly model. Achieves near-perfect performance on microbial growth media prediction (MRR: 0.9966, Precision@1: 0.9932, Hit@10: 1.0000) while maintaining explainability through attention-based reasoning over biological pathways. Training Data: - KG-Microbe merged knowledge graph: 1,379,337 nodes, 2,960,472 edges - 24 biological relation types including taxonomic hierarchies, metabolic interactions, phenotype associations, and environmental relationships - Primary prediction task: Growth media suitability for microbial taxa (biolink:occurs_in, 50K edges) - Multi-relation capability: Predicts links for any of the 24 relation types, including chemical consumption/production, phenotype associations, and taxonomic classification Citation: Original RelGT Architecture: Dwivedi et al., "Relational Graph Transformer", arXiv:2505.10960, 2025 KOGUT Implementation: Knowledge Oriented Graph Unified Transformer for Microbial Growth Media Prediction Developed at Lawrence Berkeley National Laboratory (LBNL) Trained on NERSC Perlmutter supercomputer

Joachimiak, Marcin [Lawrence Berkeley National Lab↗

NASA NTRS2001

An Assessment of the Ability of Potential Space-Borne Instruments to Resolve Spatial and Temporal Variability of Atmospheric Carbon Dioxide

Mounting concern regarding the possibility that increasing carbon dioxide concentrations will initiate climate change has stimulated interest in the feasibility of measuring CO2 mixing ratios from satellites. Currently, the most comprehensive set of atmospheric CO2 data is from the NOAA CMDL cooperative air sampling network, consisting of more than 40 sites where flasks of air are collected approximately weekly. Sporadic observations in the troposphere and stratosphere from airborne in situ and flask samplers are also available. Although the surface network is extensive, there is a dearth of data in the Southern Hemisphere and most of the stations were intentionally placed in remote areas, far from major sources. Sufficiently precise satellite observations with adequate spatial and temporal resolution would substantially increase our knowledge of the atmospheric CO2 distribution and would undoubtedly lead to improved understanding of the global carbon budget. We use a 3-D chemical transport model to investigate the ability of potential satellite instruments with a variety of orbits, horizontal resolution and vertical weighting functions to capture the variation in the modeled CO2 fields. The model is driven by analyzed winds from the Goddard Data Assimilation Office. Simulated CO2 fields are compared with existing surface and aircraft data, and the effects of the model convection scheme and representation of the planetary boundary layer are considered.

Andrews, Arlyn E.↗

NASA NTRS2001

An Assessment of the Ability of Potential Spaceborne Instruments to Resolve Spatial and Temporal Variability of Atmospheric Carbon Dioxide

Mounting concern regarding the possibility that increasing carbon dioxide concentrations will initiate climate change has stimulated interest in the feasibility of measuring CO2 mixing ratios from satellites. Currently, the most comprehensive set of atmospheric CO2 data is from the NOAA CMDL cooperative air sampling network, consisting of more than 40 sites where flasks of air are collected approximately weekly. Sporadic observations in the troposphere and stratosphere from airborne in situ and flask samplers are also available. Although the surface network is extensive, there is a dearth of data in the Southern Hemisphere and most of the stations were intentionally placed in remote areas, far from major sources. Sufficiently precise satellite observations with adequate spatial and temporal resolution would substantially increase our knowledge of the atmospheric CO2 distribution and would undoubtedly lead to improved understanding of the global carbon budget. We use a 3-D chemical transport model to investigate the ability of potential satellite instruments with a variety of orbits, horizontal resolution and vertical weighting functions to capture the variation in the modeled CO2 fields. The model is driven by analyzed winds from the Goddard Data Assimilation Office. Simulated CO2 fields are compared with existing surface and aircraft data, and the effects of the model convection scheme and representation of the planetary boundary layer are considered.

Andrews, Arlyn E.↗

NASA NTRS2006

NASA'S SERVIR Gulf of Mexico Project: The Gulf of Mexico Regional Collaborative (GoMRC)

The Gulf of Mexico Regional Collaborative (GoMRC) is a NASA-funded project that has as its goal to develop an integrated, working, prototype IT infrastructure for Earth science data, knowledge and models for the five Gulf U.S. states and Mexico, and to demonstrate its ability to help decision-makers better understand critical Gulf-scale issues. Within this preview, the mission of this project is to provide cross cutting solution network and rapid prototyping capability for the Gulf of Mexico region, in order to demonstrate substantial, collaborative, multi-agency research and transitional capabilities using unique NASA data sets and models to address regional problems. SERVIR Mesoamerica is seen as an excellent existing framework that can be used to integrate observational and GIs data bases, provide a sensor web interface, visualization and interactive analysis tools, archival functions, data dissemination and product generation within a Rapid Prototyping concept to assist decision-makers in better understanding Gulf-scale environmental issues.

Quattrochi, Dale A.↗

NASA NTRS2019

Soil Moisture Data Assimilation

Accurate knowledge of soil moisture at the continental scale is important for improving predictions of weather, agricultural productivity and natural hazards, but observations of soil moisture at such scales are limited to indirect measurements, either obtained through satellite remote sensing or from meteorological networks. Land surface models simulate soil moisture processes, using observation-based meteorological forcing data, and auxiliary information about soil, terrain and vegetation characteristics. Enhanced estimates of soil moisture and other land surface variables, along with their uncertainty, can be obtained by assimilating observations of soil moisture into land surface models. These assimilation results are of direct relevance for the initialization of hydro-meteorological ensemble forecasting systems. The success of the assimilation depends on the choice of the assimilation technique, the nature of the model and the assimilated observations, and, most importantly, the characterization of model and observation error. Systematic differences between satellite-based microwave observations or satellite-retrieved soil moisture and their simulated counterparts require special attention. Other challenges include inferring root-zone soil moisture information from observations that pertain to a shallow surface soil layer, propagating information to unobserved areas and downscaling of coarse information to finer-scale soil moisture estimates. This chapter summarizes state-of-the-art solutions to these issues with conceptual data assimilation examples, using techniques ranging from simplified optimal interpolation to spatial ensemble Kalman filtering. In addition, operational soil moisture assimilation systems are discussed that support numerical weather prediction at ECMWF and provide value-added soil moisture products for the NASA Soil Moisture Active Passive mission.

radar backscatter↗

NASA NTRS2023

The NASA Astrobiology Program’s Professional Advancement Workshop Series: Next Steps

Introduction: In the modern-day competitive job search, it is not enough for a candidate to have an excellent academic resume. Candidates must also present themselves well in various styles of interviews, have managerial skills, and be capable of engaging in complex interpersonal relationships and team building. The NASA Astrobiology Program’s Professional Advancement Workshop Series (PAWS) addresses the need for professional skills development that is not traditionally taught in graduate programs. PAWS is designed to supplement academic teachings by providing a space where early career scientists can learn new skills to help them explore, interview for, and be hired in the jobs and careers they want. This is especially true for jobs that are outside of the traditional academic “pipeline” (the path of student to postdoc to tenure-track professor). PAWS embraces the braided river model of career development [1], which emphasizes flexibility as a scientist’s needs change throughout their life. PAWS also presents an opportunity for other early career scientists to meet and network with each other outside of conferences and other formal events. This is especially important as the world emerges from the ongoing COVID-19 pandemic. Finally, PAWS creates an informative space full of resources available to everyone, which lowers institutional barriers regarding knowledge of opportunities and potentially provides mentorship to those in need. Both items are identified as supportive of a more diverse and inclusive future workforce [2–4]. PAWS began in August 2021 and was originally intended to last until the Astrobiology Science Conference in May 2022. The NASA Astrobiology Program initially sponsored PAWS, allowing the PAWS Leadership to bring in expert speakers from outside of NASA. These external speakers exposed the event attendees to new techniques and perspectives. PAWS has been able to continue beyond May 2022 because of many people volunteering their time for panels and a good balance between panels and expert-led workshops. Here, we share how PAWS works, what the results have been so far, and the future plans for PAWS.

Worshop↗