Search NASA⌕ Search

SEARCH · Search NASA

Results for “statistical learning”

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 271 records · Page 15

SigTime: Learning and Visually Explaining Time Series Signatures

Understanding and distinguishing temporal patterns in time series data is essential for scientific discovery and decision-making. For example, in biomedical research, uncovering meaningful patterns in physiological signals can improve diagnosis, risk assessment, and patient outcomes. However, existing methods for time series pattern discovery face major challenges, including high computational complexity, limited interpretability, and difficulty in capturing meaningful temporal structures. Here, to address these gaps, we introduce a novel learning framework that jointly trains two Transformer models using complementary time series representations: shapelet-based representations to capture localized temporal structures and traditional feature engineering to encode statistical properties. The learned shapelets serve as interpretable signatures that differentiate time series across classification labels. Additionally, we develop a visual analytics system—SigTime—with coordinated views to facilitate exploration of time series signatures from multiple perspectives, aiding in useful insights generation. We quantitatively evaluate our learning framework on eight publicly available datasets and one proprietary clinical dataset. Additionally, we demonstrate the effectiveness of our system through two usage scenarios along with the domain experts: one involving public ECG data and the other focused on preterm labor analysis.

97 MATHEMATICS AND COMPUTING↗

A high-throughput approach for statistical process optimization in Laser Powder Bed Fusion

Process variability is inherent in metal additive manufacturing (AM). However, it is often overlooked in process optimization frameworks, constraining the understanding of process uncertainties and their influence on parameter selection. To address this, we present an integrated framework that combines high-throughput single-track experiments, GAN-based melt pool geometry extraction, robust statistical and machine learning modeling, and uncertainty-quantified process mapping. Process variability is characterized through single-track melt pool behaviors, and its influence on defect formation is systematically quantified to enable statistically guided process parameter optimization. This approach is demonstrated on Laser Powder Bed Fusion (L-PBF) of stainless steel 316L, effectively capturing the interplay between process parameters, melt pool variability, and defect probability. By integrating uncertainty quantification into process optimization, this study provides a structured methodology for addressing variability challenges in AM quality control, ultimately contributing to enhanced manufacturing reliability.

Laser Powder Bed Fusion↗

Federal Automotive Statistical Tool: Asset-Level Data - Lessons Learned and Looking Forward

This presentation provides an overview of collecting federal motor vehicle fleet data at the individual vehicle level for the first time during the fall of 2017, lessons learned related how that data was validated, and changes being made based on those lessons. This presentation is one of three comprising a panel discussion with representatives from other Federal agencies. FAST is a Web-based information system managed by the US Department of Energy, the US General Services Administration, and the Energy Information Administration. FAST is used to collect information about the fleet of motor vehicles used and managed by the Federal government.

99 GENERAL AND MISCELLANEOUS↗

Trelliscope: A System for Detailed Visualization in Analysis of Large Complex Data

Visualization plays a critical role in the statistical model building and data analysis process. Data analysts, well-versed in statistical and machine learning methods, visualize data to hypothesize and validate models. These analysts need flexible, scalable visualization tools that are not decoupled from their analysis environment. In this paper we introduce Trelliscope, a visualization framework for statistical analysis of large complex data. Trelliscope extends Trellis, an effective visualization framework that divides data into subsets and applies a plotting method to each subset, arranging the results in rows and columns of panels. Trelliscope provides a way to create, arrange and interactively view panels for very large datasets, enabling flexible detailed visualization for data of any size. Scalability is achieved using distributed computing technologies coupled with . We discuss the underlying principles, design, and scalable architecture of Trelliscope, and illustrate its use on three analysis projects in the domains of proteomics, high energy physics, and power systems engineering.

Hafen, Ryan P.↗

Uncertainties in Density and Simulation Parameters for Radiographic Reconstructions Using Machine Learning

We will develop new and improved methods that will enable a fuller and more extensive use of experimental data, such as from radiographic imaging, towards better characterizing and reducing uncertainties in predictive modeling of weapons performance. We will achieve this by leveraging recent developments in the areas of computational imaging, statistical and machine learning, and reduced order modeling. Key new aspects of our approach include (a) a coupling of deep learning based density reconstruction with fast hydrodynamics simulators that will permit the development of certain guarantees on extrapolatability and bounds on certain other hydrodynamics-related para- metric uncertainties and (b) a dual approach towards better characterizing uncertainties in density reconstruction using deep learning based surrogates to accelerate Bayesian reconstructions on the one hand and using variational inference and Bayesian machine learning on the other hand.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

Automatized convergence of optoelectronic simulations using active machine learning

A fundamental problem of optoelectronic simulations is to achieve convergence. We use statistical analysis and machine learning to effectively guide the selection of the next device to be examined based upon the expected convergence of the simulation. This active learning strategy rapidly constructs a model that predicts Poisson-Schrödinger simulations of devices and that simultaneously produces fully converged simulations.

Rouet-Leduc, Bertrand (ORCID:0000000227917949)↗

Machine learning for energy-water nexus: challenges and opportunities

Modeling the interactions of water and energy systems is important to the enforcement of infrastructure security and system sustainability. To this end, recent technological advancement has allowed the production of large volumes of data associated with functioning of these sectors. We are beginning to see that statistical and machine learning techniques can help elucidate characteristic patterns across these systems from water availability, transport, and use to energy generation, fuel supply, and customer demand, and in the interdependencies among these systems that can leave these systems vulnerable to cascading impacts from single disruptions. In this paper, we discuss ways in which data and machine learning can be applied to the challenges facing the energy-water nexus along with the potential issues associated with the machine learning techniques themselves. We then survey machine learning techniques that have found application to date in energy-water nexus problems. We conclude by outlining future research directions and opportunities for collaboration among the energy-water nexus and machine learning communities that can lead to mutual synergistic advantage.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Machine Learning Algorithms for Matching Theories, Simulations, and Observations in Cosmology (Final Project)

The search for simple and elegant equations to explain and predict observable natural phenomena has been a driving force in scientific theory for several centuries. The great successes of Newton's laws, Maxwell's equations, the Schrodinger equation, and the theory of General Relativity have encouraged this perspective, and perhaps a beautiful equation comprising a “theory of everything” remains to be discovered. But now scientists are grappling with systems of ever-increasing complexity, fraught with nonlinearity, and with the quantities of interest often only indirectly observed. Nature may resist a simple description, and the most important discoveries of the next century may be complex theories with countless variables and parameters. The era of big data opens up a promising new approach to scientific discovery in this setting. The predictions of modern theories, even if complex or nonlinear, can be examined through detailed and computationally-intensive simulations that take days or weeks to run and fill petabytes of disk space. The upcoming generation of scientific instruments will provide petabytes of observations at unprecedented resolution and depth. The challenge is to compare data to simulations so as to test theories and identify those that best match the observations. The data are numerous and rich but are subject to noise and various systematic biases. The simulations are large and detailed but are so costly that the effective sample size is small. Moreover, the mapping from data to simulation to the parameters of interest is typically complicated and ill-conditioned. Although these challenges, and the solutions we propose, apply across the sciences, we will ground our efforts by focusing on cosmology. Here, we are truly entering the era of big data in several senses. Ever larger experiments survey larger and larger cosmological volumes, resulting in enormous amounts of pixel data to process and extract the information of cosmological interest. These include large imaging surveys, which collect all the light in broad wavelength ranges over significant fractions of the sky, and spectroscopic surveys, which measure the light as a function of wavelength to map out 3D distances. These experiments are designed to shed light on the big unanswered questions in cosmology. Distilling the petabytes of data into observed quantities for comparison with theory is a daunting task. But these large experiments do not only result in large amounts of data to analyze: if we wish to compare to theories of physics, then we need increasingly extensive predictions for the observed quantities as well, which requires ever-larger cosmological simulations that take substantial -- and at times prohibitive -- computational resources to produce. Our goal is to develop statistical and machine learning methods for using observed and simulated data to advance machine learning with applications cosmology. In particular, we focus our research efforts on the following tasks: The challenge for automated science is that it is computationally impossible and statistically dangerous to consider every possible model in order to find the best one. We will develop Bayesian Optimization based active-learning methods that accelerate both the execution of the simulations and the search for best-fitting parameters. The key idea is to make the simulations adaptive -- across resolutions, time, and parameters -- using the data to search as the simulation runs; Most ML methods operate on simple finite dimensional feature vectors. However, many cosmology and other science applications require ML methods that can operate on more complex objects as inputs or outputs such as functions, distributions, or sets and point clouds. Our goal is to develop efficient ML methods for this problem and demonstrate their applicability in Cosmology, Astrophysics, and other science problems.

79 ASTRONOMY AND ASTROPHYSICS↗

Watch and learn—a generalized approach for transferrable learning in deep neural networks via physical principles

Transfer learning refers to the use of knowledge gained while solving a machine learning task and applying it to the solution of a closely related problem. Such an approach has enabled scientific breakthroughs in computer vision and natural language processing where the weights learned in state-of-the-art models can be used to initialize models for other tasks which dramatically improve their performance and save computational time. Here we demonstrate an unsupervised learning approach augmented with basic physical principles that achieves fully transferrable learning for problems in statistical physics across different physical regimes. By coupling a sequence model based on a recurrent neural network to an extensive deep neural network, we are able to learn the equilibrium probability distributions and inter-particle interaction models of classical statistical mechanical systems. Our approach, distribution-consistent learning, DCL, is a general strategy that works for a variety of canonical statistical mechanical models (Ising and Potts) as well as disordered interaction potentials. Using data collected from a single set of observation conditions, DCL successfully extrapolates across all temperatures, thermodynamic phases, and can be applied to different length-scales. This constitutes a fully transferrable physics-based learning in a generalizable approach.

97 MATHEMATICS AND COMPUTING↗

Scalable statistical inference of photometric redshift via data subsampling

Handling big data has largely been a major bottleneck in traditional statistical models. Consequently, when accurate point prediction is the primary target, machine learning models are often preferred over their statistical counterparts for bigger problems. But full probabilistic statistical models often outperform other models in quantifying uncertainties associated with model predictions. We develop a data-driven statistical modeling framework that combines the uncertainties from an ensemble of statistical models learned on smaller subsets of data carefully chosen to account for imbalances in the input space. We demonstrate this method on a photometric redshift estimation problem in cosmology, which seeks to infer a distribution of the redshift—the stretching effect in observing the light of far-away galaxies—given multivariate color information observed for an object in the sky. Our proposed method performs balanced partitioning, graph-based data subsampling across the partitions, and training of an ensemble of Gaussian process models.

data subsampling↗

High-Throughput Discovery Illuminates Design Principles and Limits for Long-Lived Charged Species in Organic Electrolytes

The chemical stability of charged molecules in all-organic redox flow batteries (RFBs) is required for the prolonged operation of these devices. Molecular engineering and electrolyte optimization are used to mitigate parasitic reactions and extend the lifetimes of the charge carriers. However, how much can structural variation extend the lifetime? To probe this query, we designed a high-throughput kinetic study of the radical cation of N-methylphenothiazinium, guided by statistical sampling and learning algorithms. Using Argonne’s autonomous discovery facility, we conducted over 6,000 kinetic experiments with robotic sample preparation, parallel kinetic measurements, and machine learning inputs, testing 188 solvent molecules selected from a space of over 540 candidates from 11 chemical classes. Algorithmic selections guided us to stable solvent candidates, which were further tested in high concentration with and without supporting electrolyte. Our findings reveal the inherent difficulty of exceeding the current state of the art through solvent variation. The desired stability is statistically rare and poorly predictable. Among the many tested, only three solvents significantly outperformed our baseline, acetonitrile─and none by more than a factor of 3─suggesting a general challenge in achieving the necessary techno-economic targets. Furthermore, we suggest that self-discharge through solvent homolysis is the cause of the observed limitations. Several structural motifs contribute to >1,000 h half-life stability including molecular simplicity, symmetry, oxidation complement, and strategic fluorination. Importantly, this workflow establishes effective assays for diagnosing and predicting oxidative stress for highly stable liquid electrolytes in all batteries.

Batteries↗

Cluster expansion by transfer learning for phase stability predictions

Recent progress towards universal machine-learned interatomic potentials holds considerable promise for materials discovery. Yet the accuracy of these potentials for predicting phase stability may still be limited. In contrast, cluster expansions provide accurate phase stability predictions but are computationally demanding to parameterize from first principles, especially for structures of low dimension or with a large number of components, such as interfaces or multimetal catalysts. We overcome this trade-off via transfer learning. Using Bayesian inference, we incorporate prior statistical knowledge from machine-learned and physics-based potentials, enabling us to sample the most informative configurations and to efficiently fit first-principles cluster expansions. Furthermore, this algorithm is tested on Pt:Ni, showing robust convergence of the mixing energies as a function of sample size with reduced statistical fluctuations.

36 MATERIALS SCIENCE↗

Correlations between thermal history and keyhole porosity in laser powder bed fusion

Additive manufacturing has the potential to revolutionize the production of metallic components as it yields near net shape parts with complex geometries and minimizes waste. At the present day, additively manufactured components face qualification and certification challenges due to the difficulty in controlling defects. This has driven a significant research effort aimed at better understanding and improving processing controls – yielding a plethora of in-situ measurements aimed at correlating defects with material quality metrics of interest. In this work, we develop machine-learning methods to learn correlations between thermal history and subsurface porosity for a variety of print conditions in laser powder bed fusion. Un-normalized surface temperatures (in the form of black-body radiances) are obtained using high-speed infrared imaging and porosity formation is observed in the sample cross-section through synchrotron x-ray imaging. To demonstrate the predictive power of these features, we present four statistical machine-learning models that correlate temperature histories to subsurface porosity formation in laser fused Ti-6Al4V powder.

36 MATERIALS SCIENCE↗

From Organized High-Throughput Data to Phenomenological Theory using Machine Learning: The Example of Dielectric Breakdown

Understanding the behavior (and failure) of dielectric insulators experiencing extreme electric fields is critical to the operation of present and emerging electrical and electronic devices. Despite its importance, the development of a predictive theory of dielectric breakdown has remained a challenge, owing to the complex multiscale nature of this process. We focus on the intrinsic dielectric breakdown field of insulators—the theoretical limit of breakdown determined purely by the chemistry of the material, i.e., the elements the material is composed of, the atomic-level structure, and the bonding. Starting from a benchmark dataset (generated from laborious first principles computations) of the intrinsic dielectric breakdown field of a variety of model insulators, simple predictive phenomenological models of dielectric breakdown are distilled using advanced statistical or machine learning schemes, revealing key correlations and analytical relationships between the breakdown field and easily accessible material properties. Lastly, the models are shown to be general, and can hence guide the screening and systematic identification of high electric field tolerant materials.

36 MATERIALS SCIENCE↗

Low responsiveness of machine learning models to critical or deteriorating health conditions

Machine learning (ML) based mortality prediction models can be immensely useful in intensive care units. Such a model should generate warnings to alert physicians when a patient’s condition rapidly deteriorates, or their vitals are in highly abnormal ranges. Before clinical deployment, it is important to comprehensively assess a model’s ability to recognize critical patient conditions. We develop multiple medical ML testing approaches, including a gradient ascent method and neural activation map. We systematically assess these machine learning models’ ability to respond to serious medical conditions using additional test cases, some of which are time series. Guided by medical doctors, our evaluation involves multiple machine learning models, resampling techniques, and four datasets for two clinical prediction tasks. We identify serious deficiencies in the models’ responsiveness, with the models being unable to recognize severely impaired medical conditions or rapidly deteriorating health. For in-hospital mortality prediction, the models tested using our synthesized cases fail to recognize 66% of the injuries. In some instances, the models fail to generate adequate mortality risk scores for all test cases. Our study identifies similar kinds of deficiencies in the responsiveness of 5-year breast and lung cancer prediction models. Using generated test cases, we find that statistical machine-learning models trained solely from patient data are grossly insufficient and have many dangerous blind spots. Most of the ML models tested fail to respond adequately to critically ill patients. How to incorporate medical knowledge into clinical machine learning models is an important future research direction.

60 APPLIED LIFE SCIENCES↗

SU-D-BRB-01: A Comparison of Learning Methods for Knowledge Based Dose Prediction for Coplanar and Non-Coplanar Liver Radiotherapy

Purpose: The predictive power of knowledge based planning (KBP) has considerable potential in the development of automated treatment planning. Here, we examine the predictive capabilities and accuracy of previously reported KBP methods, as well as an artificial neural networks (ANN) method. Furthermore, we compare the predictive accuracy of these methods on coplanar volumetric-modulated arc therapy (VMAT) and non-coplanar 4π radiotherapy. Methods: 30 liver SBRT patients previously treated using coplanar VMAT were selected for this study. The patients were re-planned using 4π radiotherapy, which involves 20 optimally selected non-coplanar IMRT fields. ANNs were used to incorporate enhanced geometric information including liver and PTV size, prescription dose, patient girth, and proximity to beams. The performance of ANN was compared to three methods from statistical voxel dose learning (SVDL), wherein the doses of voxels sharing the same distance to the PTV are approximated by either taking the median of the distribution, non-parametric fitting, or skew-normal fitting. These three methods were shown to be capable of predicting DVH, but only median approximation can predict 3D dose. Prediction methods were tested using leave-one-out cross-validation tests and evaluated using residual sum of squares (RSS) for DVH and 3D dose predictions. Results: DVH prediction using non-parametric fitting had the lowest average RSS with 0.1176(4π) and 0.1633(VMAT), compared to 0.4879(4π) and 1.8744(VMAT) RSS for ANN. 3D dose prediction with median approximation had lower RSS with 12.02(4π) and 29.22(VMAT), compared to 27.95(4π) and 130.9(VMAT) for ANN. Conclusion: Paradoxically, although the ANNs included geometric features in addition to the distances to the PTV, it did not perform better in predicting DVH or 3D dose compared to simpler, faster methods based on the distances alone. The study further confirms that the prediction of 4π non-coplanar plans were more accurate than VMAT. NIH R43CA183390 and R01CA188300.

60 APPLIED LIFE SCIENCES↗

Analyzing petabytes of data with Hadoop

The open source Apache Hadoop project provides a powerful suite of tools for storing and analyzing petabytes of data using commodity hardware. After several years of production use inside of web companies like Yahoo! and Facebook and nearly a year of commercial support and development by Cloudera, the technology is spreading rapidly through other disciplines, from financial services and government to life sciences and high energy physics. The talk will motivate the design of Hadoop and discuss some key implementation details in depth. It will also cover the major subprojects in the Hadoop ecosystem, go over some example applications, highlight best practices for deploying Hadoop in your environment, discuss plans for the future of the technology, and provide pointers to the many resources available for learning more. In addition to providing more information about the Hadoop platform, a major goal of this talk is to begin a dialogue with the ATLAS research team on how the tools commonly used in their environment compare to Hadoop, and how Hadoop could improve better to serve the high energy physics community. Short Biography: Jeff Hammerbacher is Vice President of Products and Chief Scientist at Cloudera. Jeff was an Entrepreneur in Residence at Accel Partners immediately prior to founding Cloudera. Before Accel, he conceived, built, and led the Data team at Facebook. The Data team was responsible for driving many of the applications of statistics and machine learning at Facebook, as well as building out the infrastructure to support these tasks for massive data sets. The team produced two open source projects: Hive, a system for offline analysis built above Hadoop, and Cassandra, a structured storage system on a P2P network. Before joining Facebook, Jeff was a quantitative analyst on Wall Street. Jeff earned his Bachelor's Degree in Mathematics from Harvard University and recently served as contributing editor to the book "Beautiful Data", published by O'Reilly in July 2009.

97 MATHEMATICS AND COMPUTING↗