SEARCH · Search NASA
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.
SAND2022-1748 O Romans is a library for implementing compression-based machine learning algorithms and algorithms for Human Constrained Machine Learning. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.
The research team developed models of Attentional Control (AC) that are unique to existing modeling approaches in the literature. The goal was to enable the research team to (1) make predictions about AC and human performance in real-world scenarios and (2) to make predictions about individual characteristics based on human data. First, the team developed a proof-of-concept approach for representing an experimental design and human subjects data in a Bayesian model, then demonstrated an ability to draw inferences about conditions of interest relevant to real-world scenarios. Ultimately, this effort was successful, and we were able to make reasonable (meaning supported by behavioral data) inferences about conditions of interest to develop a risk model for AC (where risk is defined as a mismatch between AC and attentional demand). The team additionally defined a path forward for a human-constrained machine learning (HCML) approach to make predictions about an individual's state based on performance data. The effort represents a successful first step in both modeling efforts and serves as a basis for future work activities. Numerous opportunities for future work have been defined.
Volume of distribution at steady state (V D,ss ) is one of the key pharmacokinetic parameters estimated during the drug discovery process. Despite considerable efforts to predict V D,ss , accuracy and choice of prediction methods remain a challenge, with evaluations constrained to a small set (<150) of compounds. To address these issues, a series of in silico methods for predicting human V D,ss directly from structure were evaluated using a large set of clinical compounds. Machine learning (ML) models were built to predict V D,ss directly and to predict input parameters required for mechanistic and empirical V D,ss predictions. In addition, log D, fraction unbound in plasma (fup), and blood-to-plasma partition ratio (BPR) were measured on 254 compounds to estimate the impact of measured data on predictive performance of mechanistic models. Furthermore, the impact of novel methodologies such as measuring partition (Kp) in adipocytes and myocytes (n = 189) on V D,ss predictions was also investigated. In predicting V D,ss directly from chemical structures, both mechanistic and empirical scaling using a combination of predicted rat and dog V D,ss demonstrated comparable performance (62%–71% within 3-fold). The direct ML model outperformed other in silico methods (75% within 3-fold, r 2 = 0.5, AAFE = 2.2) when built from a larger data set. Scaling to human from predicted V D,ss of either rat or dog yielded poor results (<47% within 3-fold). Measured fup and BPR improved performance of mechanistic V D,ss predictions significantly (81% within 3-fold, r 2 = 0.6, AAFE = 2.0). Adipocyte intracellular Kp showed good correlation to the V D,ss but was limited in estimating the compounds with low V D,ss .
We report the design/discovery of new materials is highly nontrivial owing to the near-infinite possibilities of material candidates and multiple required property/performance objectives. Thus, machine learning tools are now commonly employed to virtually screen material candidates with desired properties by learning a theoretical mapping from material-to-property space, referred to as the forward problem. However, this approach is inefficient and severely constrained by the candidates that the human imagination can conceive. Thus, in this work on polymers, we tackle the materials discovery challenge by solving the inverse problem: directly generating candidates that satisfy desired property/performance objectives. We utilize syntax-directed variational autoencoders (VAE) in tandem with Gaussian process regression (GPR) models to discover polymers expected to be robust under three extreme conditions: (1) high temperatures, (2) high electric field, and (3) high temperature and high electric field, useful for critical structural, electrical, and energy storage applications. This approach to learn from and augment) human ingenuity is general and can be extended to discover polymers with other targeted properties and performance measures.
Machine learning techniques have received growing attention as an alternative strategy for developing general-purpose density functional approximations, augmenting the historically successful approach of human-designed functionals derived to obey mathematical constraints known for the exact exchange-correlation functional. More recently, efforts have been made to reconcile the two techniques, integrating machine learning and exact-constraint satisfaction. We continue this integrated approach, designing a deep neural network that exploits the exact constraint and appropriate norm philosophy to de-orbitalize the strongly constrained and appropriately normed (SCAN) functional. The deep neural network is trained to replicate the SCAN functional from only electron density and local derivative information, avoiding the use of the orbital-dependent kinetic energy density. The performance and transferability of the machine-learned functional are demonstrated for molecular and periodic systems.
Machine learning (ML) is transforming materials research, yet potential for biopolymer discovery remains constrained by fragmented data and nonstandardized reporting. Biopolymers differ significantly from synthetic polymers, requiring specialized approaches to represent their biosynthetic origins, hierarchical structures, and application-specific metrics. In this Perspective, we identify three core challenges limiting biopolymer representation: information encoding, data quality, and data sharing. We describe the most pressing issues and propose commensurate approaches to address each key challenge. Recommendations include the design and adoption of biopolymer-specific fingerprinting and representation frameworks, development of hybrid human-large language model (LLM) data extraction strategies, and expanding Findable, Accessible, Interoperable, Reusable (FAIR)-compliant repositories. We propose a robust foundation to define interoperable, high-quality data sets that capture the full context of biopolymer materials. Standardized metadata, shared ontologies, and community-driven infrastructure would enable scalable, reproducible workflows and accelerate the ML-driven development of biopolymers.
The fusion of experimental automation and machine learning has catalyzed a new era in materials research, prominently featuring Gaussian Process (GP) Bayesian Optimization (BO) driven autonomous experiments. Here we introduce a Dual-GP approach that enhances traditional GPBO by adding a secondary surrogate model to dynamically constrain the experimental space based on real-time assessments of the raw experimental data. This Dual-GP approach enhances the optimization efficiency of traditional GPBO by isolating more promising space for BO sampling and more valuable experimental data for primary GP training. We also incorporate a flexible, human-in-the-loop intervention method in the Dual-GP workflow to adjust for unanticipated results. We demonstrate the effectiveness of the Dual-GP model with synthetic model data and implement this approach in autonomous pulsed laser deposition experimental data. This Dual-GP approach has broad applicability in diverse GPBO-driven experimental settings, providing a more adaptable and precise framework for refining autonomous experimentation for more efficient optimization.
Quantifying the radioactive sources present in gamma spectra is an ever-present and growing national security mission and a time-consuming process for human analysts. While machine learning models exist that are trained to estimate radioisotope proportions in gamma spectra, few address the eventual need to provide explanatory outputs beyond the estimation task. In this work, we develop two machine learning models for a NaI detector measurements: one to perform the estimation task, and the other to characterize the first model’s ability to provide reasonable estimates. To ensure the first model exhibits a behavior that can be characterized by the second model, the first model is trained using a custom, semi-supervised loss function which constrains proportion estimates to be explainable in terms of a spectral reconstruction. The second auxiliary model is an out-of-distribution detection function (a type of meta-model) leveraging the proportion estimates of the first model to identify when a spectrum is sufficiently unique from the training domain and thus is out-of-scope for the model. In demonstrating the efficacy of this approach, we encourage the use of meta-models to better explain ML outputs used in radiation detection and increase trust.
Over the last 15 years, compressive sensing techniques have been developed which have the potential to greatly reduce the amount of data collected by systems while preserving the amount of information obtained. A cost of this efficiency is that a computationally-intensive optimization routine must be used to put the sensed data into a form that a person can interpret. At the same time, machine learning techniques have experienced tremendous growth as well. Machines have demonstrated the ability learn how to effectively perform tasks such as detection and classification at speeds much faster than humanly possible. Our goal in this project was to study the feasibility of using compressive sensing systems "at the edge." That is, how can compressive sensing sensors be deployed such that information is created at the remote sensor rather than sending raw data to a central processing location? Studies were performed to analyze whether machine learning could be done on the compressively sensed data in its raw form. If a machine is performing the task, is it possible to do so without putting the data into a human interpretable form? We show that this is possible for some systems, in particular a compressive sensing snapshot imaging spectrometer. Machine learning tasks were demonstrated to be more effective and more robust to noise when the machine learning algorithm worked on data in its raw form. This system is shown to outperform a traditional spectrometer. Techniques for reducing the complexity of the reconstruction routine were also analyzed. Techniques for such as data regularization, deep neural networks, and matrix completion were studied and shown to have benefits over traditional reconstruction techniques. In this project we showed that compressive sensing sensors are indeed feasible at the edge. As always, sensors and algorithms must be carefully tuned to work in the constrained environment. In this project we developed tools and techniques to enable those analyses.
Clinical deep phenotyping and phenotype annotation play a critical role in both the diagnosis of patients with rare disorders as well as in building computationally-tractable knowledge in the rare disorders field. These processes rely on using ontology concepts, often from the Human Phenotype Ontology, in conjunction with a phenotype concept recognition task (supported usually by machine learning methods) to curate patient profiles or existing scientific literature. With the significant shift in the use of large language models (LLMs) for most NLP tasks, we examine the performance of the latest Generative Pre-trained Transformer (GPT) models underpinning ChatGPT as a foundation for the tasks of clinical phenotyping and phenotype annotation. The experimental setup of the study included seven prompts of various levels of specificity, two GPT models (gpt-3.5-turbo and gpt-4.0) and two established gold standard corpora for phenotype recognition, one consisting of publication abstracts and the other clinical observations. The best run, using in-context learning, achieved 0.58 document-level F1 score on publication abstracts and 0.75 document-level F1 score on clinical observations, as well as a mention-level F1 score of 0.7, which surpasses the current best in class tool. Without in-context learning, however, performance is significantly below the existing approaches. Our experiments show that gpt-4.0 surpasses the state of the art performance if the task is constrained to a subset of the target ontology where there is prior knowledge of the terms that are expected to be matched. While the results are promising, the non-deterministic nature of the outcomes, the high cost and the lack of concordance between different runs using the same prompt and input make the use of these LLMs challenging for this particular task.
Radiation exposure poses a significant threat to human health. Emerging research indicates that even low-dose radiation once believed to be safe, may have harmful effects. This perception has spurred a growing interest in investigating the potential risks associated with low-dose radiation exposure across various scenarios. To comprehensively explore the health consequences of low-dose radiation, our study employs a robust statistical framework that examines whether specific groups of genes, belonging to known pathways, exhibit coordinated expression patterns that align with the radiation levels. Notably, our findings reveal the existence of intricate yet consistent signatures that reflect the molecular response to radiation exposure, distinguishing between low-dose and high-dose radiation. Moreover, we leverage a pathway-constrained variational autoencoder to capture the nonlinear interactions within gene expression data. By comparing these two analytical approaches, our study aims to gain valuable insights into the impact of low-dose radiation on gene expression patterns, identify pathways that are differentially affected, and harness the potential of machine learning to uncover hidden activity within biological networks. This comparative analysis contributes to a deeper understanding of the molecular consequences of low-dose radiation exposure.
To date, disinformation research has focused largely on the production of false information ignoring the suppression of select information. We term this alternative form of disinformation information suppression. Information suppression occurs when facts are withheld with the intent to mislead. In order to detect information suppression, we focus on understanding the actors who withhold information. In this research, we use knowledge of human behavior to find signatures of different gatekeeping behaviors found in text. Specifically, we build a model to classify the different types of edits on Wikipedia using the added text alone and compare a human-informed feature engineering approach to a featureless algorithm. Being able to computationally distinguish gatekeeping behaviors is a first step towards identifying when information suppression is occurring.
This dissertation describes efforts towards understanding the Higgs boson at the highest energies humanly accessible, using the CMS experiment at the Large Hadron Collider and advances in artificial intelligence (AI) and machine learning (ML). We present searches for resonant and nonresonant Higgs-boson (H) pair production in the all-hadronic two beauty-quark and two vector boson (V) final state, using a novel strategy to measure the quartic HHVV coupling and search for new Higgs-like bosons. By targeting highly Lorentz-boosted Higgs pairs, we probe effects of potential new physics in the high energy Higgs sector, which could hold answers to fundamental mysteries of nature such as baryon asymmetry. To enable these and future searches, we introduce as well significant developments in AI/ML, including in the identification of boosted H$\rightarrow$VV decays with deep transformer networks and advances in AI-accelerated fast simulations of the CMS detector. The latter notably includes the development of the first, highly performant generative models for point-cloud data in high energy physics, which have the potential to improve CMS' computational efficiency by up to three orders of magnitude. We also highlight novel solutions to the important and challenging problems of calibrating and validating these ML techniques. Finally, we present new approaches to search for new physics in a model-agnostic manner, using physics-informed ML methods equivariant to Lorentz transformations. The quartic HHVV coupling is observed (expected) to be constrained to $[-0.04, 2.05]$ ($[0.05, 1.98]$) at the 95% confidence level relative to the standard model prediction, representing the second-most sensitive measurement of this coupling by CMS to date. Exclusion limits on the production cross section of new heavy resonances decaying to two Higgs-like bosons are expected to be as low as 0.3 fb for high resonance masses.
Information from differential nuclear-physics experiments and theory is often too uncertain to accurately define nuclear-physics observables such as cross sections or energy spectra. Integral experimental data, representing the applications of these observables, are often more precise but depend simultaneously on too many of them to unambiguously identify issues in the observable with human expert analysis alone. Here, we explore how we can leverage physics knowledge gained from differential experimental data, nuclear theory, integral experiments, and neutron-transport calculations to better understand nuclear-physics observables in the context of the application area represented by integral experiments. We support this task with machine-learning methods to discern trends in a large amount of convoluted data. Differential and integral information was used in an analysis augmented by the random forest and the Shapley additive explanations metric. We chose as an application area one that is represented by criticality measurements and pulsed-sphere neutron-leakage spectra. We show one representative example ( 241 Pu fission observables) where the combination of differential and integral information allowed to resolve issues in data representing these observables. As a starting point, the machine learning (ML) algorithms highlighted several observables as leading potentially to bias in simulating integral experiments. Differential information, paired with sensitivity to integral quantities, allowed us then to pinpoint one specific observable ( 241 Pu fission cross section) as the main driver of bias. The comparison to integral experiments, on the other hand, allowed us to indicate a likely reliable experiment among several discrepant ones for this observables. In other cases (e.g., 239 Pu observables), we were not able to resolve the confounding introduced by integral experiments but instead highlighted the need for targeted new experiments and theory developments to better constrain the nuclear-physics space for the application area represented by integral experiments. We were able to combine information from differential experimental data, nuclear-physics theory, integral experiments, and neutron-transport simulations of the latter experiments with the help of the random forest algorithm and expert judgment. This combination of knowledge allows to improve our description of nuclear-physics observables as applied to a particular application area.
Watersheds regulate biogeochemical processes and provide ecosystem services to human societies, but disturbances can fundamentally alter these processes across space and time. Determining when and where to sample to capture disturbance impacts in watersheds remains a central challenge. Manipulation studies and long-term monitoring are often constrained by scope, and opportunistic studies often lack pre-disturbance data needed to statistically determine disturbance impacts. We identify a persistent knowledge gap: the absence of a clear, transferable framework to guide opportunistic disturbance research where pre-disturbance data collection is not a feasible option. To address this gap, we present a conceptual framework that intentionally integrates modeling and empirical observation in an iterative, stepwise model–experiment workflow. We demonstrate its application through two contrasting case studies: wildfire impacts on headwater streams using a pre-disturbance preparedness approach, and saltwater flooding impacts on coastal forests using an ‘ex-post-facto’ approach. From these applications, we assess strengths, limitations, and the critical role of team science for transferability across disturbance types and study designs. Broadly, this framework offers a scalable path towards more rigorous, timely, and actionable disturbance science that can inform watershed management, hazard risk reduction, and ecosystem resilience.