Search NASA⌕ Search

SEARCH · Search NASA

Results for “Domain knowledge”

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 199 records · Page 11

Uncertainty-refined image segmentation under domain shift

Digital image segmentation is provided. The method comprises training a neural network for image segmentation with a labeled training dataset from a first domain, wherein a subset of nodes in the neural net are dropped out during training. The neural network receives image data from a second, different domain. A vector of N values that sum to 1 is calculated for each image element, wherein each value represents an image segmentation class. A label is assigned to each image element according to the class with the highest value in the vector. Multiple inferences are performed with active dropout layers for each image element, and an uncertainty value is generated for each image element. Uncertainty is resolved according to expected characteristics. The label of any image element with an uncertainty above a threshold is replaced with a new label corresponding to a segmentation class based on domain knowledge.

Martinez, Carianne↗

CAD-VAE: Leveraging Correlation-Aware Latents for Comprehensive Fair Disentanglement

While deep generative models have significantly advanced representation learning, they may inherit or amplify biases and fairness issues by encoding sensitive attributes alongside predictive features. Enforcing strict independence in disentanglement is often unrealistic when target and sensitive factors are naturally correlated. To address this challenge, we propose CAD-VAE(Correlation-Aware Disentangled VAE), which introduces a correlated latent code to capture the information shared between the target and sensitive attributes. Given this correlated latent, our method effectively separates over-lapping factors without extra domain knowledge by directly minimizing the conditional mutual information between target and sensitive codes. A relevance-driven optimization strategy refines the correlated code by efficiently capturing essential correlated features and eliminating redundancy. Extensive experiments on benchmark datasets demonstrate that CAD-VAE produces fairer representations, realistic counterfactuals, and improved fairness-aware image editing.

Ma, Chenrui [University of California Irvine]↗

NOUS: Construction and Querying of Dynamic Knowledge Graphs

The ability to construct domain specific knowledge graphs (KG) and perform question-answering or hypothesis generation is a trans- formative capability. Despite their value, automated construction of knowledge graphs remains an expensive technical challenge that is beyond the reach for most enterprises and academic institutions. We propose an end-to-end framework for developing custom knowl- edge graph driven analytics for arbitrary application domains. The uniqueness of our system lies A) in its combination of curated KGs along with knowledge extracted from unstructured text, B) support for advanced trending and explanatory questions on a dynamic KG, and C) the ability to answer queries where the answer is embedded across multiple data sources.

Choudhury, Sutanay↗

Velo: A Knowledge Management Framework for Modeling and Simulation

Modern scientific enterprises are inherently knowledge-intensive. Scientific studies in domains such as geosciences, climate, and biology require the acquisition and manipulation of large amounts of experimental and field data to create inputs for large-scale computational simulations. The results of these simulations are then analyzed, leading to refinements of inputs and models and additional simulations. The results of this process must be managed and archived to provide justifications for regulatory decisions and publications that are based on the models. In this paper we introduce our Velo framework that is designed as a reusable, domain independent knowledge management infrastructure for modeling and simulation. Velo leverages, integrates and extends open source collaborative and content management technologies to create a scalable and flexible core platform that can be tailored to specific scientific domains. We describe the architecture of Velo for managing and associating the various types of data that are used and created in modeling and simulation projects, as well as the framework for integrating domain-specific tools. To demonstrate realizations of Velo, we describe examples from two deployed sites for carbon sequestration and climate modeling. These provide concrete example of the inherent extensibility and utility of our approach.

99 GENERAL AND MISCELLANEOUS↗

Velo: Riding the Knowledge Management Wave for Simulation and Modeling

Modern scientific enterprises are inherently knowledge-intensive. In general, scientific studies in domains such as geosciences, climate, and biology require the acquisition and manipulation of large amounts of experimental and field data in order to create inputs for large-scale computational simulations. The results of these simulations must then be analyzed, leading to refinements of inputs and models and further simulations. Further, these results must be managed and archived to provide justifications for publications and regulatory decisions that are based on these models. In this paper we describe our Velo framework that is designed as a reusable, domain independent knowledge management infrastructure for modeling and simulation. Velo leverages, integrates, and extends open source collaborative and content management technologies to create a scalable and flexible core platform that can be tailored to specific scientific domains. In this paper we describe the architecture of Velo for managing and associating the various types of data that are used and created in modeling and simulation projects, as well as the framework for integrating domain-specific tools. To demonstrate a realization of Velo, we describe the Geologic Sequestration Software Suite (GS3) that has been developed to support geologic sequestration modeling. This provides a concrete example of the inherent extensibility and utility of our approach.

99 GENERAL AND MISCELLANEOUS↗

LLMs for Mfg.—On the State of Large Language Models and Applications to Manufacturing

Additive Manufacturing (AM), referred to as 3D printing, has emerged as a key pillar of Industry 4.0 enabling layer-by-layer fabrication of intricate geometries from CAD models. In parallel, Large Language Models (LLMs), deep learning models for natural language generation trained on vast text corpora, have demonstrated unprecedented capabilities in understanding and generating human-like text. The convergence of these trends opens new opportunities at the intersection of AM and AI/ML, where LLMs can assist engineers and researchers in design, manufacture planning, and knowledge discovery. Recent academic work has begun to explore LLM applications in AM and adjacent fields, such as material science, mechanical engineering, and design for additive manufacturing. This exploration ranges from intelligent process planning to domain-specific knowledge retrieval. This survey provides a comprehensive review of current developments, focusing on peer-reviewed literature contributions that apply, adapt, and advance LLMs in general and domain-specific domains. We analyze state-of-the-art (SOTA) techniques, such as fine-tuning foundational models for specific domains, retrieval-augmented generation (RAG) pipelines, knowledge graph integration, and delve into the architectures and evaluation methods employed. The goal of this survey is to inform researchers and practitioners of the current capabilities and limitations of LLMs in general and in domain-specific applications, and to outline how these models are being tailored to meet the requirements of these applications.

36 MATERIALS SCIENCE↗

KEBLM: Knowledge-Enhanced Biomedical Language Models

Pretrained language models (PLMs) have demonstrated strong performance on many natural language processing (NLP) tasks. Despite their great success, these PLMs are typically pretrained only on unstructured free texts without leveraging existing structured knowledge bases that are readily available for many domains, especially scientific domains. As a result, these PLMs may not achieve satisfactory performance on knowledge-intensive tasks such as biomedical NLP. Comprehending a complex biomedical document without domain-specific knowledge is challenging, even for humans. Inspired by this observation, we propose a general framework for incorporating various types of domain knowledge from multiple sources into biomedical PLMs. We encode domain knowledge using lightweight adapter modules, bottleneck feed-forward networks that are inserted into different locations of a backbone PLM. For each knowledge source of interest, we pretrain an adapter module to capture the knowledge in a self-supervised way. We design a wide range of self-supervised objectives to accommodate diverse types of knowledge, ranging from entity relations to description sentences. Once a set of pretrained adapters is available, we employ fusion layers to combine the knowledge encoded within these adapters for downstream tasks. Each fusion layer is a parameterized mixer of the available trained adapters that can identify and activate the most useful adapters for a given input. Our method diverges from prior work by including a knowledge consolidation phase, during which we teach the fusion layers to effectively combine knowledge from both the original PLM and newly-acquired external knowledge using a large collection of unannotated texts. After the consolidation phase, the complete knowledge-enhanced model can be fine-tuned for any downstream task of interest to achieve optimal performance. Extensive experiments on many biomedical NLP datasets show that our proposed framework consistently improves the performance of the underlying PLMs on various downstream tasks such as natural language inference, question answering, and entity linking. These results demonstrate the benefits of using multiple sources of external knowledge to enhance PLMs and the effectiveness of the framework for incorporating knowledge into PLMs. Finally, while primarily focused on the biomedical domain in this work, our framework is highly adaptable and can be easily applied to other domains, such as the bioenergy sector.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

GraphAide: Advanced Graph-Assisted Query and Reasoning System

Curating knowledge from multiple siloed sources that contain both structured and unstructured data is a major challenge in many real-world applications. Pattern matching and querying represent fundamental tasks in modern data analytics that leverage this curated knowledge. The development of such applications necessitates overcoming several research challenges, including data extraction, named entity recognition, data modeling, and designing query interfaces. Moreover, the explainability of these functionalities is critical for their broader adoption. The emergence of Large Language Models (LLMs) has accelerated the development lifecycle of new capabilities. Nonetheless, there is an ongoing need for domain-specific tools tailored to user activities. The creation of digital assistants has gained considerable traction in recent years, with LLMs offering a promising avenue to develop such assistants utilizing domain-specific knowledge and assumptions. In this context, we introduce an advanced query and reasoning system, GraphAide, which constructs a knowledge graph (KG) from diverse sources and allows to query and reason over the resulting KG. GraphAide harnesses both the KG and LLMs to rapidly develop domain-specific digital assistants. It integrates design patterns from retrieval augmented generation (RAG) and the semantic web to create an agentic LLM application. GraphAide underscores the potential for streamlined and efficient development of specialized digital assistants, thereby enhancing their applicability across various domains.

Purohit, Sumit [BATTELLE (PACIFIC NW LAB)] (ORCID:↗

Science Literacy Project, August 2006 - August 2008

Ball State University (BSU) was the recipient of a U.S. Department of Energy award to develop educational games teaching science and math. The Science Media Program will merge Ball State University’s nationally recognized capabilities in education, technology, and communication to develop new, interactive, game-based media for the teaching and learning of science and scientific principles for K-12 students. BSU established a team of educators, researchers, scientists, animators, designers, technology specialists, and hired a professional media developer company (Outside Source Design) from Indianapolis. After six months discussions and assessments the project team selected the following 8 games in Math, Physics, Chemistry, and Biology, 2 from each discipline. The assembled teams were innovative and unique. This new model of development and production included a process that integrated all needed knowledge and expertise for the development of high quality science and math games for K-12 students. This new model has potential to be used by others for the development of the educational games. The uniqueness of the model is to integrate domain experts’ knowledge with researchers/quality control group, and combine a professional development team from the game development company with the academic game development team from Computer Science and Art departments at Ball State University. The developed games went through feasibility tests with selected students for improvement before use in the research activities.

59 BASIC BIOLOGICAL SCIENCES↗

Solving the puzzle of Fe homeostasis by integrating molecular, mathematical, and societal models

To ensure optimal utilization and bioavailability, iron uptake, transport, subcellular localization, and assimilation are tightly regulated in plants. In this work, we examine recent advances in our understanding of cellular responses to Fe deficiency. We then use intracellular mechanisms of Fe homeostasis to discuss how formalizing cell biology knowledge via a mathematical model can advance discovery even when quantitative data is limited. Using simulation-based inference to identify plausible systems mechanisms that conform to known emergent phenotypes can yield novel, testable hypotheses to guide targeted experiments. However, this approach relies on the accurate encoding of domain-expert knowledge in exploratory mathematical models. We argue that this would be facilitated by fostering more “systems thinking” life scientists and that diversifying your research team may be a practical path to achieve that goal.

59 BASIC BIOLOGICAL SCIENCES↗

Experimental mapping of soluble protein domains using a hierarchical approach

Exploring the function and 3D space of large multidomain protein targets often requires sophisticated experimentation to obtain the targets in a form suitable for structure determination. Screening methods capable of selecting well-expressed, soluble fragments from DNA libraries exist, but require the use of automation to maximize chances of picking a few good candidates. Here, we describe the use of an insertion dihydrofolate reductase (DHFR) vector to select in-frame fragments and a split-GFP assay technology to filter-out constructs that express insoluble protein fragments. With the incorporation of an IPCR step to create high density, focused sublibraries of fragments, this cost-effective method can be performed manually with no a priori knowledge of domain boundaries while permitting single amino acid resolution boundary mapping. We used it on the well-characterized p85a subunit of the phosphoinositide-3-kinase to demonstrate the robustness and efficiency of our methodology. We then successfully tested it onto the polyketide synthase PpsC from Mycobacterium tuberculosis, a potential drug target involved in the biosynthesis of complex lipids in the cell envelope. X-ray quality crystals from the acyl-transferase (AT), dehydratase (DH) and enoyl-reductase (ER) domains have been obtained.

59 BASIC BIOLOGICAL SCIENCES↗

TU-F-BRD-01: Biomedical Informatics for Medical Physicists

Biomedical informatics encompasses a very large domain of knowledge and applications. This broad and loosely defined field can make it difficult to navigate. Physicists often are called upon to provide informatics services and/or to take part in projects involving principles of the field. The purpose of the presentations in this symposium is to help medical physicists gain some knowledge about the breadth of the field and how, in the current clinical and research environment, they can participate and contribute. Three talks have been designed to give an overview from the perspective of physicists and to provide a more in-depth discussion in two areas. One of the primary purposes, and the main subject of the first talk, is to help physicists achieve a perspective about the range of the topics and concepts that fall under the heading of 'informatics'. The approach is to de-mystify topics and jargon and to help physicists find resources in the field should they need them. The other talks explore two areas of biomedical informatics in more depth. The goal is to highlight two domains of intense current interest--databases and models--in enough depth into current approaches so that an adequate background for independent inquiry is achieved. These two areas will serve as good examples of how physicists, using informatics principles, can contribute to oncology practice and research. Learning Objectives: To understand how the principles of biomedical informatics are used by medical physicists. To put the relevant informatics concepts in perspective with regard to biomedicine in general. To use clinical database design as an example of biomedical informatics. To provide a solid background into the problems and issues of the design and use of data and databases in radiation oncology. To use modeling in the service of decision support systems as an example of modeling methods and data use. To provide a background into how uncertainty in our data and knowledge can be incorporated into modeling methods.

97 MATHEMATICS AND COMPUTING↗

CACTUS: Chemistry Agent Connecting Tool Usage to Science

Large language models (LLMs) have shown remarkable potential in various domains but often lack the ability to access and reason over domain-specific knowledge and tools. In this article, we introduce Chemistry Agent Connecting Tool-Usage to Science (CACTUS), an LLM-based agent that integrates existing cheminformatics tools to enable accurate and advanced reasoning and problem-solving in chemistry and molecular discovery. We evaluate the performance of CACTUS using a diverse set of open-source LLMs, including Gemma-7b, Falcon-7b, MPT-7b, Llama3-8b, and Mistral-7b, on a benchmark of thousands of chemistry questions. Our results demonstrate that CACTUS significantly outperforms baseline LLMs, with the Gemma-7b, Mistral-7b, and Llama3-8b models achieving the highest accuracy regardless of the prompting strategy used. Moreover, we explore the impact of domain-specific prompting and hardware configurations on model performance, highlighting the importance of prompt engineering and the potential for deploying smaller models on consumer-grade hardware without a significant loss in accuracy. By combining the cognitive capabilities of open-source LLMs with widely used domain-specific tools provided by RDKit, CACTUS can assist researchers in tasks such as molecular property prediction, similarity searching, and drug-likeness assessment.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Using Knowledge-Guided Machine Learning To Assess Patterns of Areal Change in Waterbodies across the Contiguous United States

Lake and reservoir surface areas are an important proxy for freshwater availability. Advancements in machine learning (ML) techniques and increased accessibility of remote sensing data products have enabled the analysis of waterbody surface area dynamics on broad spatial scales. However, interpreting the ML results remains a challenge. While ML provides important tools for identifying patterns, the resultant models do not include mechanisms. Thus, the “black-box” nature of ML techniques often lacks ecological meaning. Using ML, we characterized temporal patterns in lake and reservoir surface area change from 1984 to 2016 for 103,930 waterbodies in the contiguous United States. We then employed knowledge-guided machine learning (KGML) to classify all waterbodies into seven ecologically interpretable groups representing distinct patterns of surface area change over time. Many waterbodies were classified as having “no change” (43%), whereas the remaining 57% of waterbodies fell into other groups representing both linear and nonlinear patterns. This analysis demonstrates the potential of KGML not only for identifying ecologically relevant patterns of change across time but also for unraveling complex processes that underpin those changes.

58 GEOSCIENCES↗

Using Knowledge-Guided Machine Learning To Assess Patterns of Areal Change in Waterbodies across the Contiguous United States

Lake and reservoir surface areas are an important proxy for freshwater availability. Advancements in machine learning (ML) techniques and increased accessibility of remote sensing data products have enabled the analysis of waterbody surface area dynamics on broad spatial scales. However, interpreting the ML results remains a challenge. While ML provides important tools for identifying patterns, the resultant models do not include mechanisms. Thus, the “black-box” nature of ML techniques often lacks ecological meaning. Using ML, we characterized temporal patterns in lake and reservoir surface area change from 1984 to 2016 for 103,930 waterbodies in the contiguous United States. We then employed knowledge-guided machine learning (KGML) to classify all waterbodies into seven ecologically interpretable groups representing distinct patterns of surface area change over time. Many waterbodies were classified as having “no change” (43%), whereas the remaining 57% of waterbodies fell into other groups representing both linear and nonlinear patterns. This analysis demonstrates the potential of KGML not only for identifying ecologically relevant patterns of change across time but also for unraveling complex processes that underpin those changes.

54 ENVIRONMENTAL SCIENCES↗

Lifetime and Degradation Study of Poly(Methyl Methacrylate) via a Data-Driven Study Protocol Approach

We report to optimize and extend the service life of polymeric materials in outdoor environments, a domain knowledge-based and data-driven approach was utilized to quantitatively investigate the temporal evolution of degradation modes, mechanisms, and rates under various stepwise accelerated exposure conditions. Six formulations of poly(methyl methacrylate) (PMMA) with different combinations of stabilizing additives, including one unstabilized formulation, were exposed in three accelerated weathering conditions. Degradation was dependent on wavelength as samples in UV light at 340 nm (UVA) exposure showed the most yellowing. The unstabilized PMMA formulation showed much higher yellowness index values (59.5) than stabilized PMMA formulations (2–12). Urbach edge analysis shows a shift toward longer wavelength from 285 to 500 nm with increasing exposure time and an increased absorbance around 400 nm of visible region as the unstabilized samples increase in yellowing. The degradation mechanisms of PMMA were tracked using induced absorbance to dose at specific wavelengths that correspond to known degradation mechanisms. The degradation pathway of PMMA was modeled in a framework using network structural equation modeling (netSEM). netSEM showed changes in degradation pathway as PMMA transition stages of degradation.

36 MATERIALS SCIENCE↗

A data-centric weak supervised learning for highway traffic incident detection

Using the data from loop detector sensors for near-real-time detection of traffic incidents on highways is crucial to averting major traffic congestion. While recent supervised machine learning methods offer solutions to incident detection by leveraging human-labeled incident data, the false alarm rate is often too high to be used in practice. Specifically, the inconsistency in the human labeling of the incidents significantly affects the performance of supervised learning models. To that end, we focus on a data-centric approach to improve the accuracy and reduce the false alarm rate of traffic incident detection on highways. We develop a weak supervised learning workflow to generate high-quality training labels for the incident data without the ground truth labels, and we use those generated labels in the supervised learning setup for final detection. This approach comprises three stages. First, we introduce a data preprocessing and curation pipeline that processes traffic sensor data to generate high-quality training data through leveraging labeling functions, which can be domain knowledge-related or simple heuristic rules. Second, we evaluate the training data generated by weak supervision using three supervised learning models-random forest, k-nearest neighbors, and a support vector machine ensemble-and long short-term memory classifiers. The results show that the accuracy of all of the models improves significantly after using the training data generated by weak supervision. Third, we develop an online real-time incident detection approach that leverages the model ensemble and the uncertainty quantification while detecting incidents. Finally, we show that our proposed weak supervised learning workflow achieves a high incident detection rate (0.90) and low false alarm rate (0.08).

97 MATHEMATICS AND COMPUTING↗