Search NASA⌕ Search

SEARCH · Search NASA

Results for “Training Analysis”

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 973 records · Page 54

Influence of Multiple Genetic Polymorphisms on Genitourinary Morbidity After Carbon Ion Radiotherapy for Prostate Cancer

Purpose: To investigate the genetic risk of late urinary morbidity after carbon ion radiotherapy in prostate cancer patients. Methods and Materials: A total of 197 prostate cancer patients who had undergone carbon ion radiotherapy were evaluated for urinary morbidity. The distribution of patients with dysuria was as follows: Grade 0, 165; Grade 1, 28; and Grade 2, 4 patients. The patients were divided (2:1) consecutively into the training and test sets and then categorized into control (Grade 0) and case (Grade 1 or greater) groups. First, 450 single nucleotide polymorphisms (SNPs) in 118 candidate genes were genotyped in the training set. The associations between the SNP genotypes and urinary morbidity were assessed using Fisher's exact test. Then, various combinations of the markers were tested for their ability to maximize the area under the receiver operating characteristics (AUC-ROC) curve analysis results. Finally, the test set was validated for the selected markers. Results: When the SNP markers in the SART1, ID3, EPDR1, PAH, and XRCC6 genes in the training set were subjected to AUC-ROC curve analysis, the AUC-ROC curve reached a maximum of 0.86. The AUC-ROC curve of these markers in the test set was 0.77. The SNPs in these five genes were defined as 'risk genotypes.' Approximately 90% of patients in the case group (Grade 1 or greater) had three or more risk genotypes. Conclusions: Our results have shown that patients with late urinary morbidity after carbon ion radiotherapy can be stratified according to the total number of risk genotypes they harbor.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Initial Verification of GEOS-4 Aerosols Using CALIPSO and MODIS: Scene Classification

A-train sensors such as MODIS and MISR provide column aerosol properties, and in the process a means of estimating aerosol type (e.g. smoke vs. dust). Correct classification of aerosol type is important because retrievals are often dependent upon selection of the right aerosol model. In addition, aerosol scene classification helps place the retrieved products in context for comparisons and analysis with aerosol transport models. The recent addition of CALIPSO to the A-train now provides a means of classifying aerosol distribution with altitude. CALIPSO level 1 products include profiles of attenuated backscatter at 532 and 1064 nm, and depolarization at 532 nm. Backscatter intensity, wavelength ratio, and depolarization provide information on the vertical profile of aerosol concentration, size, and shape. Thus similar estimates of aerosol type using MODIS or MISR are possible with CALIPSO, and the combination of data from all sensors provides a means of 3D aerosol scene classification. The NASA Goddard Earth Observing System general circulation model and data assimilation system (GEOS-4) provides global 3D aerosol mass for sulfate, sea salt, dust, and black and organic carbon. A GEOS-4 aerosol scene classification algorithm has been developed to provide estimates of aerosol mixtures along the flight track for NASA's Geoscience Laser Altimeter System (GLAS) satellite lidar. GLAS launched in 2003 and did not have the benefit of depolarization measurements or other sensors from the A-train. Aerosol typing from GLAS data alone was not possible, and the GEOS-4 aerosol classifier has been used to identify aerosol type and improve the retrieval of GLAS products. Here we compare 3D aerosol scene classification using CALIPSO and MODIS with the GEOS-4 aerosol classifier. Dust, smoke, and pollution examples will be discussed in the context of providing an initial verification of the 3D GEOS-4 aerosol products. Prior model verification has only been attempted with surface mass comparisons and column optical depth from AERONET and MODIS.

Welton, Ellsworth J.↗

Subsonic Aircraft With Regression and Neural-Network Approximators Designed

At the NASA Glenn Research Center, NASA Langley Research Center's Flight Optimization System (FLOPS) and the design optimization testbed COMETBOARDS with regression and neural-network-analysis approximators have been coupled to obtain a preliminary aircraft design methodology. For a subsonic aircraft, the optimal design, that is the airframe-engine combination, is obtained by the simulation. The aircraft is powered by two high-bypass-ratio engines with a nominal thrust of about 35,000 lbf. It is to carry 150 passengers at a cruise speed of Mach 0.8 over a range of 3000 n mi and to operate on a 6000-ft runway. The aircraft design utilized a neural network and a regression-approximations-based analysis tool, along with a multioptimizer cascade algorithm that uses sequential linear programming, sequential quadratic programming, the method of feasible directions, and then sequential quadratic programming again. Optimal aircraft weight versus the number of design iterations is shown. The central processing unit (CPU) time to solution is given. It is shown that the regression-method-based analyzer exhibited a smoother convergence pattern than the FLOPS code. The optimum weight obtained by the approximation technique and the FLOPS code differed by 1.3 percent. Prediction by the approximation technique exhibited no error for the aircraft wing area and turbine entry temperature, whereas it was within 2 percent for most other parameters. Cascade strategy was required by FLOPS as well as the approximators. The regression method had a tendency to hug the data points, whereas the neural network exhibited a propensity to follow a mean path. The performance of the neural network and regression methods was considered adequate. It was at about the same level for small, standard, and large models with redundancy ratios (defined as the number of input-output pairs to the number of unknown coefficients) of 14, 28, and 57, respectively. In an SGI octane workstation (Silicon Graphics, Inc., Mountainview, CA), the regression training required a fraction of a CPU second, whereas neural network training was between 1 and 9 min, as given. For a single analysis cycle, the 3-sec CPU time required by the FLOPS code was reduced to milliseconds by the approximators. For design calculations, the time with the FLOPS code was 34 min. It was reduced to 2 sec with the regression method and to 4 min by the neural network technique. The performance of the regression and neural network methods was found to be satisfactory for the analysis and design optimization of the subsonic aircraft.

Patnaik, Surya N.↗

In-Transit Data Transport Strategies for Coupled AI-Simulation Workflow Patterns

Coupled AI-Simulation workflows are becoming the major workloads for HPC facilities, and their increasing complexity necessitates new tools for performance analysis and prototyping of new in-situ workflows. We present SimAI-Bench, a tool designed to both prototype and evaluate these coupled workflows. In this paper, we use SimAI-Bench to benchmark the data transport performance of two common patterns on the Aurora supercomputer: a one-to-one workflow with co-located simulation and AI training instances, and a many-to-one workflow where a single AI model is trained from an ensemble of simulations. For the one-to-one pattern, our analysis shows that node-local and DragonHPC data staging strategies provide excellent performance compared Redis and Lustre file system. For the many-to-one pattern, we find that data transport becomes a dominant bottleneck as the ensemble size grows. Our evaluation reveals that file system is the optimal solution among the tested strategies for the many-to-one pattern.

Tummalapalli, Harikrishna [Argonne National Labora↗

Fast neural network training on a cluster of GPUs for action recognition with high accuracy

In this work, we propose algorithms and techniques to accelerate training of deep neural networks for action recognition on a cluster of GPUs. The convergence analysis of our algorithm shows it is possible to reduce communication cost and at the same time minimize the number of iterations needed for convergence. We customize the Adam optimizer for our distributed algorithm to improve efficiency. In addition, we employ transfer-learning to further reduce training time while improving validation accuracy. For the UCF101 and HMDB51 datasets, the validation accuracies achieved are 93.1% and 67.9% respectively. With an additional end-to-end trained temporal stream, the validation accuracies achieved for UCF101 and HMDB51 are 93.47% and 81.24% respectively. As far as we know, these are the highest accuracies achieved with the two-stream approach using ResNet that does not involve computationally expensive 3D convolutions or pretraining on much larger datasets.

97 MATHEMATICS AND COMPUTING↗

Results of the 2013-2015 Association of Residents in Radiation Oncology Survey of Chief Residents in the United States

Purpose: The purpose of this project was to survey radiation oncology chief residents to define their residency experience and readiness for independent practice. Methods and Materials: During the academic years 2013 to 2014 and 2014 to 2015, the Association of Residents in Radiation Oncology (ARRO) conducted an electronic survey of post-graduate year-5 radiation oncology residents in the United States during the final 3 months of training. Descriptive statistics are reported. Results: Sixty-six chief residents completed the survey in 2013 to 2014 (53% response rate), and 69 completed the survey in 2014 to 2015 (64% response rate). Forty to 85% percent of residents reported inadequate exposure to high-dose rate and low-dose rate brachytherapy. Nearly all residents in both years (>90%) reported adequate clinical experience for the following disease sites: breast, central nervous system, gastrointestinal, genitourinary, head and neck, and lung. However, as few as 56% reported adequate experience in lymphoma or pediatric malignancies. More than 90% of residents had participated in retrospective research projects, with 20% conducting resident-led prospective clinical trials and 50% conducting basic science or translational projects. Most chief residents reported working 60 or fewer hours per week in the clinical/hospital setting and performing fewer than 15 hours per week tasks that were considered to have little or no educational value. There was more than 80% compliance with Accreditation Council for Graduate Medical Education (ACGME) work hour limits. Fifty-five percent of graduating residents intended to join an established private practice group, compared to 25% who headed for academia. Residents perceive the job market to be more competitive than previous years. Conclusions: This first update of the ARRO chief resident survey since the 2007 to 2008 academic year documents US radiation oncology residents' experiences and conditions over a 2-year period. This analysis may serve as a valuable tool for those seeking to improve training of the next generation of oncology leaders.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Wide‐Field Bond Quality Evaluation Using Frequency Domain Thermoreflectance with Deep Neural Network Feature Reconstruction

Heterogeneous integration of microelectronic components provides a pathway to improve circuit/component performance; however, this comes with assembly challenges, in particular due to complex interfaces via subsurface bump bonds. The ability of these bonds to transmit electrical signals and conduct heat to the carrier substrate limits component performance. In this work, hyperspectral frequency‐domain thermoreflectance (FDTR) imaging is demonstrated as a robust technique for evaluating the quality of subsurface indium bump bonds in a surrogate microelectronic sample. By performing microscale FDTR imaging with coarse motion image stitching, thermal phase maps that cover a 4 mm by 4 mm field‐of‐view with subsurface feature sensitivity at depths greater than 50 µm are obtained. The resulting FDTR hyperspectral data contains more than three million pixels and reveal the quality of subsurface microbump arrays. Wide‐field analysis of bonded versus gap regions is enabled by deep neural network feature reconstruction, that after training, rapidly provides an interpretable representation of bond quality. Utility of noisy higher frequency FDTR phase maps, i.e., near the computationally predicted sensing depth limit, results in an average prediction error of 11%. Taken together, FDTR with neural network‐based analysis demonstrates subsurface bond monitoring at length scales relevant for heterogeneously integrated microelectronics.

FDTR↗

Quantitative Imaging of Cobalt Phthalocyanine Distribution on Carbon Nanotubes: A Deep Learning Approach to Catalyst Characterization

Electrochemical reduction of carbon dioxide (CO 2 ) offers a pathway to valuable products, with catalysts playing a crucial role. This study investigates the distribution of cobalt tetraaminophthalocyanine (CoPc-NH 2 ) immobilized on carbon nanotubes (CNTs), utilizing high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) to characterize CoPc-NH 2 distribution. A challenge in the quantitative HAADF-STEM analysis is the introduction of bias from manual Co atom identification. To address this, we developed and trained a convolutional neural network (CNN) using a data set generated from images of CoPc-NH 2 /CNT samples with varying Co loadings. The CNN, implemented in TensorFlow and Keras, facilitated Co atom detections. Analysis of the CNN-generated data confirmed a correlation between Co loading and surface density, consistent with findings from UV–vis spectroscopy. Furthermore, the application of Ripley’s L(d) function highlighted the presence of slight Co atom clustering. Furthermore, this work demonstrates the utility of the combined HAADF-STEM and CNN approach for providing spatially resolved information about catalyst distribution on nonplanar supports, revealing structural details that are typically lost through other characterization methods.

HAADF-STEM↗

Verification Data Pattern Recognition and Change Detection at the Neutron Instrument Level

As the pressure on International Atomic Energy Agency (IAEA) safeguards resources increases, there is a drive to essentially “do more with less.” New ways of using current verification instrumentation and visualizing the collected data are needed, coupled with maintaining and sustaining the current safeguards instrumentation fleet. This paper explores how to innovate with current safeguards neutron instrumentation using data-driven algorithms at the instrument level. A concept is described for upgrading current neutron counting systems with modern electronics by leveraging electronics developments in other fields (e.g., high-energy physics) to facilitate full list mode data acquisition and, therefore, expand existing analyses to pattern recognition and change detection at the instrument level. The aim is to advance correlated neutron counting by improving the reliability of detecting nuclear material diversion and, at the same time, to improve the sustainability of those systems. It is likely that upgrading existing systems with new hardware and software is more cost-effective than replacing systems with other novel systems that might meet this need. A specific project example is highlighted in which a system was upgraded by retrofitting the standard uranium neutron collar detector with new electronics; as a result, verification data can be analyzed in new ways. A new neutron analysis method, the “List Mode Response Matrix,” has been developed to analyze every pulse train from the detector array and thus enable the verification of spatial information about the assay item for safeguards inspections. This project has improved the reliability of resolving individual fuel pin locations within a given fresh fuel assembly and the percentage of partial defects detected. Herein, the overall concept of upgrading current neutron counting systems is discussed with respect to the potential increase in analysis capabilities. In addition, the details and results of the list mode collar project are discussed.

Worrall, Louise G.↗

Deep learning to estimate permeability using geophysical data

Time-lapse electrical resistivity tomography (ERT) is a popular geophysical method to estimate three-dimensional (3D) permeability fields from electrical potential difference measurements. Traditional inversion and data assimilation methods are used to ingest this ERT data into hydrogeophysical models to estimate permeability. Due to ill-posedness and the curse of dimensionality, existing inversion strategies provide poor estimates and low resolution of the 3D permeability field. Recent advances in deep learning provide us with powerful algorithms to overcome this challenge. This paper presents a deep learning (DL) framework to estimate the 3D subsurface permeability from time-lapse ERT data. To test the feasibility of the proposed framework, we train DL-enabled inverse models on simulation data. Each measurement in both synthetic and field data is standardized by removing the mean and scaling the time-series to unit variance. This pre-processing step is necessary to bring simulation data closer to field observations. Subsurface process models based on hydrogeophysics are used to generate this synthetic data. Training performed on limited simulation data resulted in the DL model over-fitting. An advanced data augmentation based on mixup is implemented to generate additional training samples to overcome this issue. This mixup technique creates weakly labeled (low-fidelity) samples from strongly labeled (high-fidelity) data. The weakly labeled training data is then used to develop DL-enabled inverse models and reduce over-fitting. As both time-lapse ERT (1133048 features/realization) and 3D permeability (585453 features/realization) data samples are from a high-dimensional space, principal component analysis (PCA) is employed to reduce dimensionality. Encoded ERT and encoded permeability are generated using the trained PCA estimators. A deep neural network is then trained to map the encoded ERT to encoded permeability. This mixup training and unsupervised learning allowed us to build a fast and reasonably accurate DL-based inverse model under limited simulation data. Results show that proposed weak supervised learning can capture salient spatial features in the 3D permeability field. Quantitatively, the average mean squared error (in terms of the natural log) on the strongly labeled training, validation, and test datasets is less than 0.5. The R 2 -score (global metric) is greater than 0.75, and the percent error in each cell (local metric) is less than 10%. Finally, an added benefit in terms of computational cost is that the proposed DL-based inverse model is at least O(10 4 ) times faster than running a forward model once it is trained. Data generation, DL model training, and hyperparameter tuning to identify optimal neural network architectures utilized high-performance computing resources while the DL inference is performed on a standard laptop. Approximately, O(10 5 ) processor hours are used for generating data and DL tuning and training. We acknowledge that the data generation and DL model development are expensive. But once a DL model is trained, it can be re-used for inversion rapidly for the given system, with set physics and domain. Note that traditional inversion may require multiple forward model simulations (e.g., in the order of 10 to 1000), which are very expensive. This computational savings ≈ O(10 5 ) – O(10 7 )) makes the proposed DL-based inverse model attractive for subsurface imaging and real-time ERT monitoring applications due to fast and yet reasonably accurate estimations of permeability field.

58 GEOSCIENCES↗

Collaborative Research: Vlasov-Maxwell Simulations to Resolve Electron Heating and Dissipation, in Quasi-Perpendicular Shocks (Final Report)

Collisionless shocks are a grand challenge problem in plasma physics and have been the subject of study for more than six decades. Shocks are ubiquitous phenomena in the universe and are responsible for transforming high energy and momentum flows into thermal energy and energetic particles. Understanding how shocks operate is of primary importance to understand the sun-earth coupling, protecting manned missions and spacecrafts from high energy particles, inertial confinement fusion, and radiation observed from astrophysical plasmas, such as supernova remnants and astrophysical jets. A major unanswered question on this frontier is, how does a collisionless plasma transform flow energy and momentum into electron thermal energy? Many potential mechanisms have been proposed to perform the conversion between flow and thermal energy in a collisionless plasma, but the answer has been elusive. To solve the mystery, non-relativistic perpendicular and quasi-perpendicular shocks are studied via a coordinated approach employing cutting-edge numerical simulations with the continuum kinetic code Gkeyll and spacecraft observations supplied by NASA's Magnetospheric Multiscale (MMS) mission. In both sets of data, we employ novel techniques to identify how and where energy is exchanged between electromagnetic fields and plasma particles. By combining cutting-edge spacecraft data and numerical simulations with these diagnostics, we have been able to identify some of the primary mechanisms responsible electron heating and dissipation in heliospheric shocks. In addition to electron dynamics, we have complete access to proton simulation and in situ data, enabling us to characterize proton heating and acceleration in quasi-perpendicular shocks. Many astrophysical quasi-perpendicular shocks have similar parameters to those in this study, including supernova remnants and shocks in galaxy clusters. Additionally, the simulation code, Gkeyll, and associated analysis scripts are openly available for community use. This award has supported a collaborative effort between Princeton University, the University of Arizona - Tucson, and the University of Maryland - College Park. The award funded three early career scientists, including the funding and training of one female postdoctoral researcher, who was involved in all aspects of the MMS data analysis and some portions of the simulation analysis.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Uncertainty-Aware Machine Learning for Small-Angle X-ray Scattering Analysis in Autonomous Experimentation

Small-angle X-ray scattering (SAXS) is a powerful high-throughput characterization tool for probing nanoscale structure in native sample environments, providing real-time morphological information such as nanoparticle size and shape during synthesis. However, automated SAXS data analysis for extracting meaningful structural parameters is non-trivial and remains a bottleneck in closed-loop experimentation towards autonomous materials discovery, which demands fast, reliable, and uncertainty-aware data analysis. Here, we develop a machine-learning approach for automated SAXS analysis tailored to closed-loop nanoparticle synthesis. A Random Forest (RF) regression model is trained on 100,000 synthetic SAXS curves generated from polydisperse spherical nanoparticles with realistic background contributions. Using normalized one-dimensional SAXS intensity profiles as input, the RF model directly predicts nanoparticle radius, size polydispersity, and background parameters, while the ensemble standard deviation across trees provides built-in uncertainty quantification (UQ). On synthetic data, we show that combining fit-quality metrics (R 2 , MAE) with thresholds on prediction uncertainty reliably identifies accurate parameter estimates without access to ground truth. We then apply the trained model to 365 experimental SAXS profiles of citrate-reduced gold nanoparticles synthesized using an automated droplet-flow microreactor with in situ SAXS at a synchrotron beamline, classifying the results into high- and low-confidence subsets based on UQ metrics. Finally, we integrate RF-based SAXS analysis into a simulated closed-loop optimization campaign using Gaussian process Bayesian optimization to minimize nanoparticle polydispersity, benchmarking against conventional automated Levenberg–Marquardt fitting. The RF-guided campaign exhibits substantially faster convergence and lower relative opportunity cost (∼0.07 vs ∼0.3), demonstrating that uncertainty-aware machine-learning SAXS analysis significantly enhances the efficiency and robustness of autonomous nanomaterials synthesis workflows.

Bayesian optimization↗

Inter-Individual Variability in Astronaut Strength Response to Long-Duration Spaceflight May Be Dictated by In-Flight Exercise

BACKGROUND: Deconditioning induced by prolonged microgravity exposure during spaceflight hinders physical capacity. Examination of pre- to post-flight isokinetic strength measures shows that strength decreases; however, considerable variability between crewmembers exists. As in-flight exercise differs between crew, an examination of exercise training variables contributing to strength preservation is necessary. The purpose of this analysis was to examine in-flight exercise of astronauts who maintain strength compared to those with the greatest decrements in strength from pre- to post-flight. METHODS: The database of NASA International Space Station (ISS) crew who completed pre- and post-flight knee isokinetic testing (n = 87) was used to identify “responders” (R) and “non-responders” (NR) to in-flight exercise (≥ 90-day flight duration). R were those that maintained or improved strength (mean of knee extension and flexion at 60°/sec) (n = 14; 5.6 ± 7.8%), and NR were those with the most severe decrements (n = 14; -32.8 ± 4.1%) from pre- to post-flight. In-flight daily exercise records (cycle ergometer, treadmill, resistance exercise) were aggregated on a weekly basis. Independent t-tests or Mann-Whitney U tests were used to compare groups for in-flight exercise variables (i.e., frequency, time, intensity, volume). RESULTS: In-flight cycle ergometer frequency (2.4 ± 0.8 vs. 3.3 ± 1.0 days/wk; p = 0.021), total time (78.5 ± 20.4 vs. 114.0 ± 49.3 min/wk; p = 0.021), and active time (59.3 ± 17.2 vs. 86.9 ± 44.1 min/wk; p = 0.050) were greater in NR than R, but no differences for cycle intensity were observed. Treadmill distance traveled per session was greater in R than NR (4.6 ± 0.8 vs 3.6 ± 1.3 km/sess, p = 0.031), but no differences for treadmill frequency, total time, or active time were observed. Lower body resistance exercise intensity relative to pre-flight body weight was greater in R than NR (110.0 ± 16.5 vs 88.7 ± 25.0%; p = 0.012), but no differences in upper body or core resistance exercise intensity, volume, or frequency existed. CONCLUSIONS: The unique musculoskeletal loading provided by treadmill running at the expense of cycle ergometry, combined with using higher relative lower body loads during resistance training, may contribute to preserving strength in ISS crewmembers during long-duration spaceflight.

Alyssa N. Varanoske↗

Analysis of dashpot performance for rotating control drums of a lithium cooled fast reactor concept

A dashpot was incorporated in the design of the drive train of the rotating control drum to prevent shock damage to the control drum and drive train at the termination of a scram action. A rotating vane dashpot using reactor coolant lithium as a damping fluid appears to be the best candidate of the various damping devices explored. A performance analysis, results and discussion of vane type dashpots are presented.

Wenzler, C. J.↗

ERISTAR: Earth Resources Information Storage, Transformation, Analysis, and Retrieval administrative report

The 1972 Systems Engineering program at Marshall Space Flight Center where 15 participants representing 15 U.S. universities, 1 NASA/MSFC employee, and another specially assigned faculty member, participated in an 11-week program is discussed. The Fellows became acquainted with the philosophy of systems engineering, and as a training exercise, used this approach to produce a conceptional design for an Earth Resources Information Storage, Transformation, Analysis, and Retrieval System. The program was conducted in three phases; approximately 3 weeks were devoted to seminars, tours, and other presentations to subject the participants to technical and other aspects of the information management problem. The second phase, 5 weeks in length, consisted of evaluating alternative solutions to problems, effecting initial trade-offs and performing preliminary design studies and analyses. The last 3 weeks were occupied with final trade-off sessions, final design analyses and preparation of a final report and oral presentation.

Vachon, R. I.↗

Design of Neural Networks for Fast Convergence and Accuracy

A novel procedure for the design and training of artificial neural networks, used for rapid and efficient controls and dynamics design and analysis for flexible space systems, has been developed. Artificial neural networks are employed to provide a means of evaluating the impact of design changes rapidly. Specifically, two-layer feedforward neural networks are designed to approximate the functional relationship between the component spacecraft design changes and measures of its performance. A training algorithm, based on statistical sampling theory, is presented, which guarantees that the trained networks provide a designer-specified degree of accuracy in mapping the functional relationship. Within each iteration of this statistical-based algorithm, a sequential design algorithm is used for the design and training of the feedforward network to provide rapid convergence to the network goals. Here, at each sequence a new network is trained to minimize the error of previous network. The design algorithm attempts to avoid the local minima phenomenon that hampers the traditional network training. A numerical example is performed on a spacecraft application in order to demonstrate the feasibility of the proposed approach.

Maghami, Peiman G.↗

LTAU-FF: Loss Trajectory Analysis for Uncertainty in atomistic Force Fields

Model ensembles are effective tools for estimating prediction uncertainty in deep learning atomistic force fields. However, their widespread adoption is hindered by high computational costs and overconfident error estimates. In this work, we address these challenges by leveraging distributions of per-sample errors obtained during training and employing a distance-based similarity search in the model latent space. Our method, which we call LTAU (Loss Trajectory Analysis for Uncertainty), efficiently estimates the full probability distribution function of errors for any test point using the logged training errors, achieving speeds that are 2–3 orders of magnitudes faster than typical ensemble methods and allowing it to be used for tasks where training or evaluating multiple models would be infeasible. We apply LTAU towards estimating parametric uncertainty in atomistic force fields (LTAU-FF), demonstrating that it produces well-calibrated confidence intervals and predicts errors that correlate strongly with the true errors for data near the training domain. Furthermore, we show that the errors predicted by LTAU-FF can be used in practical applications for detecting out-of-domain data, tuning model performance, and predicting failure during simulations. We believe that LTAU will be a valuable tool for uncertainty quantification in atomistic force fields and is a promising method that should be further explored in other domains of machine learning.

97 MATHEMATICS AND COMPUTING↗

Integrating Deep Learning and Hydrodynamic Modeling to Improve the Great Lakes Forecast

The Laurentian Great Lakes, one of the world’s largest surface freshwater systems, pose a modeling challenge in seasonal forecast and climate projection. While physics-based hydrodynamic modeling is a fundamental approach, improving the forecast accuracy remains critical. In recent years, machine learning (ML) has quickly emerged in geoscience applications, but its application to the Great Lakes hydrodynamic prediction is still in its early stages. This work is the first one to explore a deep learning approach to predicting spatiotemporal distributions of the lake surface temperature (LST) in the Great Lakes. Our study shows that the Long Short-Term Memory (LSTM) neural network, trained with the limited data from hypothetical monitoring networks, can provide consistent and robust performance. The LSTM prediction captured the LST spatiotemporal variabilities across the five Great Lakes well, suggesting an effective and efficient way for monitoring network design in assisting the ML-based forecast. Furthermore, we employed an explainable artificial intelligence (XAI) technique named SHapley Additive exPlanations (SHAP) to uncover how the features impact the LSTM prediction. Our XAI analysis shows air temperature is the most influential feature for predicting LST in the trained LSTM. The relatively large bias in the LSTM prediction during the spring and fall was associated with substantial heterogeneity of air temperature during the two seasons. In contrast, the physics-based hydrodynamic model performed better in spring and fall yet exhibited relatively large biases during the summer stratification period. Finally, we developed a statistical integration of the hydrodynamic modeling and deep learning results based on the Best Linear Unbiased Estimator (BLUE). The integration further enhanced prediction accuracy, suggesting its potential for next-generation Great Lakes forecast systems.

Xue, Pengfei (ORCID:000000025702421X)↗