Search NASA⌕ Search

SEARCH · Search NASA

Results for “DATASETS”

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 1,765 records · Page 98
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Global transpiration data from sap flow measurements: the SAPFLUXNET database

Plant transpiration links physiological responses of vegetation to water supply and demand with hydrological, energy, and carbon budgets at the land–atmosphere interface. However, despite being the main land evaporative flux at the global scale, transpiration and its response to environmental drivers are currently not well constrained by observations. Here we introduce the first global compilation of whole-plant transpiration data from sap flow measurements (SAPFLUXNET, https://sapfluxnet.creaf.cat/, last access: 8 June 2021). We harmonized and quality-controlled individual datasets supplied by contributors worldwide in a semi-automatic data workflow implemented in the R programming language. Datasets include sub-daily time series of sap flow and hydrometeorological drivers for one or more growing seasons, as well as metadata on the stand characteristics, plant attributes, and technical details of the measurements. SAPFLUXNET contains 202 globally distributed datasets with sap flow time series for 2714 plants, mostly trees, of 174 species. SAPFLUXNET has a broad bioclimatic coverage, with woodland/shrubland and temperate forest biomes especially well represented (80 % of the datasets). The measurements cover a wide variety of stand structural characteristics and plant sizes. The datasets encompass the period between 1995 and 2018, with 50 % of the datasets being at least 3 years long. Accompanying radiation and vapour pressure deficit data are available for most of the datasets, while on-site soil water content is available for 56 % of the datasets. Many datasets contain data for species that make up 90 % or more of the total stand basal area, allowing the estimation of stand transpiration in diverse ecological settings. SAPFLUXNET adds to existing plant trait datasets, ecosystem flux networks, and remote sensing products to help increase our understanding of plant water use, plant responses to drought, and ecohydrological processes. SAPFLUXNET version 0.1.5 is freely available from the Zenodo repository (https://doi.org/10.5281/zenodo.3971689; Poyatos et al., 2020a). The “sapfluxnetr” R package – designed to access, visualize, and process SAPFLUXNET data – is available from CRAN.

58 GEOSCIENCES↗
NASA NTRSDate not supplied

Creating Benchmark Data for Artificial Intelligence and Machine Learning Space Biology Research

To identify an appropriate AI/ML approach for a specific problem, the best practice is to measure algorithm performance through the benchmarking process. A scientific benchmark consists of an AI-ready dataset and a reference implementation on a specific scientific question. The NASA Science Mission Directorate (SMD) has started the “Benchmark Initiative for AI/ML to create scientific benchmark datasets in three applications: 1) scientific benchmarking, which finds the best algorithm for a specific problem; 2) application benchmarking, which measures algorithm performance against a set of parameters; and 3) system benchmarking, which evaluates performance of hardware and software architecture. Currently, there are no standardized datasets available to benchmark AI/ML algorithms in the domain of space biology. In this work, we constructed two AI/ML-ready biological datasets from experiments in space-flown mice: cellular imaging and RNA-seq. First, radiation-exposed immune cells harbor DNA damage foci that can be fluorescently marked to visualize the amount of damage following exposure to ionizing radiation. However, such large datasets are difficult to analyze visually, due to imaging inconsistencies and human bias, and classical image processing approaches can fail on imaging artifacts. AI/ML are therefore exciting alternative, providing the speed of machines and the accuracy of humans. We have made this dataset available at https://registry.opendata.aws/bps_microscopy/. Second, high-throughput nucleic acid sequencing (DNA-seq, RNA-seq) has become widespread in biomedical research due to the growing availability and affordability of these assays. However, most sequencing datasets suffer from high dimensionality and low sample count. In this work, we used a generative adversarial network to synthesize a standardized, AI-ready, publicly available benchmark dataset for space biology RNA-seq data with sufficient space-flown and ground control mouse liver samples from NASA GeneLab. This dataset is available at https://registry.opendata.aws/bps_rnaseq/. These datasets are now fully open the Space Biology community to test their favorite AI/ML approaches.

James Casaletto↗
DOE OSTI2012

A comparative assessment of monthly mean wind speed products over the global ocean

The accurate estimation of marine wind speed is important for climate and air–sea interaction applications. There are many datasets of monthly mean wind speeds available based either on in situ measurements, satellite retrievals, atmospheric reanalysis assimilating both in situ and satellite data and blended datasets combining some or all of these other data sources. Twelve different monthly mean wind speed datasets are compared for the period from 1987 to 2009. The results suggest that we cannot presently be confident that the mean wind speed over the ocean is known to the accuracy required for the calculation of air–sea heat fluxes. Comparisons are complicated by different representations of wind speed being presented in different datasets. The in situ and reanalysis datasets present stability-dependent, earth-relative, wind speeds adjusted to a reference level of 10 m. The satellite and blended datasets present neutral equivalent, surface-relative, speeds adjusted to a reference level of 10 m. Differences between these estimates depend on atmospheric stability and ocean currents and can be greater than the required accuracy target. The adjustment for stability is itself uncertain but it is shown that these uncertainties are likely to be smaller than biases caused when the effects of stability are neglected. Further differences among the datasets are identified. Biases are caused by unidentified rain in Ku-band scatterometer-derived wind speeds and by atmospheric effects on passive microwave wind retrievals. When satellite observations affected by rain are removed a fair-weather bias remains. Some datasets are biased low in coastal regions by the effects of lower wind speeds over land in atmospheric models affecting wind speeds near the coast. All these uncertainties combine to give a wide range of estimates of monthly mean wind speed for the chosen datasets with uncertainty in mean values, spatial patterns and changes over time.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

scPlantAnnotate: an accurate and robust transformer-based model for plant cell type annotation

Accurate cell type annotation remains a major bottleneck in plant single-cell RNA sequencing (scRNA-seq), where existing tools are often adapted from animal studies and perform sub-optimally on plant data. The lack of plant-specific computational frameworks limits the construction of plant cell atlases and downstream biological discovery. We develop and evaluate scPlantAnnotate, a Transformer-based reference annotation framework tailored for plant scRNA-seq data, and benchmark it against state-of-the-art deep learning and conventional methods across multiple plant species. Species-specific scPlantAnnotate models were trained using curated datasets from Arabidopsis thaliana, Zea mays, Oryza sativa, and Glycine max. We compared scPlantAnnotate with leading baselines under both standard random-split evaluation and a more stringent leave-one-dataset-out setting, which tests robustness to completely unseen datasets and tissue types. scPlantAnnotate consistently outperforms existing approaches across all four species under random-split evaluation. In the leave-one-dataset-out setting for A. thaliana, where performance drops markedly for all methods due to strong batch effects and dataset heterogeneity, scPlantAnnotate nonetheless achieves the highest Accuracy, Macro-F1, Balanced Accuracy, and Macro-AUROC on average and ranks first on most held-out datasets. These results demonstrate improved robustness to dataset shifts, a critical yet underexplored challenge in plant scRNA-seq analysis. A freely accessible web server enables users to annotate their own datasets using pretrained models. scPlantAnnotate provides a plant-specific, Transformer-based framework for single-cell annotation that delivers state-of-the-art performance and enhanced robustness to unseen datasets. By addressing limitations of existing tools and enabling scalable reference-based annotation, scPlantAnnotate supports the development of comprehensive plant cell atlases and facilitates broader use of single-cell genomics in plant biology.

Bioinformatics↗
DOE OSTI2023

DeepAstroUDA: semi-supervised universal domain adaptation for cross-survey galaxy morphology classification and anomaly detection

Abstract Artificial intelligence methods show great promise in increasing the quality and speed of work with large astronomical datasets, but the high complexity of these methods leads to the extraction of dataset-specific, non-robust features. Therefore, such methods do not generalize well across multiple datasets. We present a universal domain adaptation method, DeepAstroUDA , as an approach to overcome this challenge. This algorithm performs semi-supervised domain adaptation (DA) and can be applied to datasets with different data distributions and class overlaps. Non-overlapping classes can be present in any of the two datasets (the labeled source domain, or the unlabeled target domain), and the method can even be used in the presence of unknown classes. We apply our method to three examples of galaxy morphology classification tasks of different complexities (three-class and ten-class problems), with anomaly detection: (1) datasets created after different numbers of observing years from a single survey (Legacy Survey of Space and Time mock data of one and ten years of observations); (2) data from different surveys (Sloan Digital Sky Survey (SDSS) and DECaLS); and (3) data from observing fields with different depths within one survey (wide field and Stripe 82 deep field of SDSS). For the first time, we demonstrate the successful use of DA between very discrepant observational datasets. DeepAstroUDA is capable of bridging the gap between two astronomical surveys, increasing classification accuracy in both domains (up to 40 % on the unlabeled data), and making model performance consistent across datasets. Furthermore, our method also performs well as an anomaly detection algorithm and successfully clusters unknown class samples even in the unlabeled target dataset.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Unpaired image translation to mitigate domain shift in liquid argon time projection chamber detector responses

Deep learning algorithms often are developed and trained on a training dataset and deployed on test datasets. Any systematic difference between the training and a test dataset may severely degrade the final algorithm performance on the test dataset—what is known as the domain shift problem . This issue is prevalent in many scientific domains where algorithms are trained on simulated data but applied to real-world datasets. Typically, the domain shift problem is solved through various domain adaptation (DA) methods. However, these methods are often tailored for a specific downstream task, such as classification or semantic segmentation, and may not easily generalize to different tasks. This work explores the feasibility of using an alternative way to solve the domain shift problem that is not specific to any downstream algorithm. The proposed approach relies on modern Unpaired Image-to-Image (UI2I) translation techniques, designed to find translations between different image domains in a fully unsupervised fashion. In this study, the approach is applied to a domain shift problem commonly encountered in Liquid Argon Time Projection Chamber (LArTPC) detector research when seeking a way to translate samples between two differently distributed LArTPC detector datasets deterministically. This translation allows for mapping real-world data into the simulated data domain where the downstream algorithms can be run with much less domain-shift-related performance degradation. Conversely, using the translation from the simulated data to a real-world domain can increase the realism of the simulated dataset and reduce the magnitude of any systematic uncertainties. To evaluate the quality of the translations, we use both pixel-wise metrics and a downstream task to measure the effectiveness of UI2I methods for mitigating the domain shift problem. We adapted several popular UI2I translation algorithms to work on scientific data and demonstrated the viability of these techniques for solving the domain shift problem with LArTPC detector data. To facilitate further development of DA techniques for scientific datasets, the ‘Simple Liquid-Argon Track Samples’ dataset used in this study is also published.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Training a Quantum Annealing Based Restricted Boltzmann Machine on Cybersecurity Data

A restricted Boltzmann machine (RBM) is a generative model that could be used in effectively balancing a cybersecurity dataset because the synthetic data a RBM generates follows the probability distribution of the training data. RBM training can be performed using contrastive divergence (CD) and quantum annealing (QA). QA-based RBM training is fundamentally different from CD and requires samples from a quantum computer. We present a real-world application that uses a quantum computer. Specifically, we train a RBM using QA for cybersecurity applications. The D-Wave 2000Q has been used to implement QA. RBMs are trained on the ISCX data, which is a benchmark dataset for cybersecurity. For comparison, RBMs are also trained using CD. CD is a commonly used method for RBM training. Our analysis of the ISCX data shows that the dataset is imbalanced. We present two different schemes to balance the training dataset before feeding it to a classifier. The first scheme is based on the undersampling of benign instances. The imbalanced training dataset is divided into five sub-datasets that are trained separately. A majority voting is then performed to get the result. Our results show the majority vote increases the classification accuracy up from 90.24% to 95.68%, in the case of CD. For the case of QA, the classification accuracy increases from 74.14% to 80.04%. In the second scheme, a RBM is used to generate synthetic data to balance the training dataset. We show that both QA and CD-trained RBM can be used to generate useful synthetic data. Balanced training data is used to evaluate several classifiers. Among the classifiers investigated, K-Nearest Neighbor (KNN) and Neural Network (NN) perform better than other classifiers. They both show an accuracy of 93%. Our results show a proof-of-concept that a QA-based RBM can be trained on a 64-bit binary dataset. The illustrative example suggests the possibility to migrate many practical classification problems to QA-based techniques. Further, we show that synthetic data generated from a RBM can be used to balance the original dataset.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2017

Evolving Deep Networks Using HPC

While a large number of deep learning networks have been studiedand published that produce outstanding results on natural imagedatasets, these datasets only make up a fraction of those to whichdeep learning can be applied. These datasets include text data, audiodata, and arrays of sensors that have very different characteristicsthan natural images. As these “best” networks for natural imageshave been largely discovered through experimentation and cannotbe proven optimal on some theoretical basis, there is no reasonto believe that they are the optimal network for these drasticallydifferent datasets. Hyperparameter search is thus often a very im-portant process when applying deep learning to a new problem. Inthis work we present an evolutionary approach to searching thepossible space of network hyperparameters and construction thatcan scale to 18, 000 nodes. This approach is applied to datasets ofvarying types and characteristics where we demonstrate the abilityto rapidly find best hyperparameters in order to enable practitionersto quickly iterate between idea and result.

Young, Steven↗
DOE OSTI2022

Overview and How-to Tutorial Videos for Using NEWTS Data

Overview and How-to Tutorial Videos for Using NEWTS Data Video 1: Overview of NEWTS Database Video 2: How-to tutorial for EPA Flue Gas Desulfurization (FGD) Effluent NEWTS dataset Video 3: How-to tutorial for USGS Produced Waters NEWTS dataset Video 4: How-to tutorial for EPA Ash NEWTS dataset Video 5: How-to tutorial for Quillinan, et al 2018 DOE Geothermal Technology Office REE dataset Video 6: Tutorial video on navigating the NEWTS Dashboard, with an overview of NEWTS and navigating between the NEWTS Dashboard and Datasets (https://netl-doe.maps.arcgis.com/apps/dashboards/a5fa4192f7c6478dab3d6180d9c30b84) Video 7: Additional tutorial video on navigating the NEWTS Dashboard and investigating specific data points in the Dashboard and Datasets Video 8: Re-record of recent webinar giving an overview of the NEWTS Database and Dashboard, including interacting with the NEWTS Dashboard, locating specific data points, and finding the relevant streams in the NEWTS Database and datasets on EDX. Includes overview of the datasets, case studies, and steps for taking stream data from the database and modeling stream data in OLI Studio and Geochemist's Workbench. Note: Video 3 tutorial is also applicable to the USGS Brackish Water NEWTS dataset.

Aqueous Chemistry↗
DOE OSTI2014

Commercial and Residential Hourly Load Profiles for all TMY3 Locations in the United States

Note: This dataset has been superseded by the dataset found at "End-Use Load Profiles for the U.S. Building Stock" (submission 4520; linked in the submission resources), which is a comprehensive and validated representation of hourly load profiles in the U.S. commercial and residential building stock. The End-Use Load Profiles project website includes links to data viewers for this new dataset. For documentation of dataset validation, model calibration, and uncertainty quantification, see Wilson et al. (2022). These data were first created around 2012 as a byproduct of various analyses of solar photovoltaics and solar water heating (see references below for are two examples). This dataset contains several errors and limitations. It is recommended that users of this dataset transition to the updated version of the dataset posted in the resources. This dataset contains weather data, commercial load profile data, and residential load profile data. Weather The Typical Meteorological Year 3 (TMY3) provides one year of hourly data for around 1,000 locations. The TMY weather represents 30-year normals, which are typical weather conditions over a 30-year period. Commercial The commercial load profiles included are the 16 ASHRAE 90.1-2004 DOE Commercial Prototype Models simulated in all TMY3 locations, with building insulation levels changing based on ASHRAE 90.1-2004 requirements in each climate zone. The folder names within each resource represent the weather station location of the profiles, whereas the file names represent the building type and the representative city for the ASHRAE climate zone that was used to determine code compliance insulation levels. As indicated by the file names, all building models represent construction that complied with the ASHRAE 90.1-2004 building energy code requirements. No older or newer vintages of buildings are represented. Residential The BASE residential load profiles are five EnergyPlus models (one per climate region) representing 2009 IECC construction single-family detached homes simulated in all TMY3 locations. No older or newer vintages of buildings are represented. Each of the five climate regions include only one heating fuel type; electric heating is only found in the Hot-Humid climate. Air conditioning is not found in the Marine climate region. One major issue with the residential profiles is that for each of the five climate zones, certain location-specific algorithms from one city were applied to entire climate zones. For example, in the Hot-Humid files, the heating season calculated for Tampa, FL (December 1 - March 31) was unknowingly applied to all other locations in the Hot-Humid zone, which restricts heating operation outside of those days (for example, heating is disabled in Dallas, TX during cold weather in November). This causes the heating energy to be artificially low in colder parts of that climate zone, and conversely the cooling season restriction leads to artificially low cooling energy use in hotter parts of each climate zone. Additionally, the ground temperatures for the representative city were used across the entire climate zone. This affects water heating energy use (because inlet cold water temperature depends on ground temperature) and heating/cooling energy use (because of ground heat transfer through foundation walls and floors). Representative cities were Tampa, FL (Hot-Humid), El Paso, TX (Mixed-Dry/Hot-Dry), Memphis, TN (Mixed-Humid), Arcata, CA (Marine), and Billings, MT (Cold/Very-Cold). The residential dataset includes a HIGH building load profile that was intended to provide a rough approximation of older home vintages, but it combines poor thermal insulation with larger house size, tighter thermostat setpoints, and less efficient HVAC equipment. Conversely, the LOW building combines excellent thermal insulation with smaller house size, wider thermostat setpoints, and more efficient HVAC equipment. However, it is not known how well these ...

Array↗
DOE OSTI2019

On the Use of NLDAS2 Weather Data for Hydrologic Modeling in the Upper Mississippi River Basin

Weather data are the key forces that drive hydrological processes so that their accuracy in watershed modeling is fundamentally important. For large-scale watershed modeling, weather data are either generated by using interpolation methods or derived from assimilated datasets. In the present study, we compared model performances of the Soil and Water Assessment Tool (SWAT), as driven by interpolation weather data, and NASA North American Land Data Assimilation System Phase Two (NLDAS2) weather dataset in the Upper Mississippi River Basin (UMRB). The SWAT model fed with different weather datasets were used to simulate monthly stream flow at 11 United States Geological Survey (USGS) monitoring stations in the UMRB. Model performances were evaluated based on three metrics: coefficient of determination (R 2 ), Nash–Sutcliffe coefficient (NS), and percent bias (P bias ). The results show that, after calibration, the SWAT model compared well at all monitoring stations for monthly stream flow using different weather datasets indicating that the SWAT model can adequately produce long-term water yield in UMRB. The results also show that using NLDAS2 weather dataset can improve SWAT prediction of monthly stream flow with less prediction uncertainty in the UMRB. We concluded that NLDAS2 dataset could be used by the SWAT model for large-scale watersheds like UMRB as a surrogate of the interpolation weather data. Further analyses results show that NLDAS2 daily solar radiation data was about 2.5 MJ m -2 higher than the interpolation data. As such, the SWAT model driven by NLDAS2 dataset tended to underestimate stream flow in the UMRB due to the overestimation in evapotranspiration in uncalibrated conditions. Thus, the implication of overestimated solar radiation by NLDAS2 dataset should be considered before using NLDAS2 dataset to drive the hydrological model.

54 ENVIRONMENTAL SCIENCES↗
NASA NTRS2017

Security Vulnerability Profiles of NASA Mission Software: Empirical Analysis of Security Related Bug Reports

NASA develops, runs, and maintains software systems for which security is of vital importance. Therefore, it is becoming an imperative to develop secure systems and extend the current software assurance capabilities to cover information assurance and cybersecurity concerns of NASA missions. The results presented in this report are based on the information provided in the issue tracking systems of one ground mission and one flight mission. The extracted data were used to create three datasets: Ground mission IVV issues, Flight mission IVV issues, and Flight mission Developers issues. In each dataset, we identified the software bugs that are security related and classified them in specific security classes. This information was then used to create the security vulnerability profiles (i.e., to determine how, why, where, and when the security vulnerabilities were introduced) and explore the existence of common trends. The main findings of our work include:- Code related security issues dominated both the Ground and Flight mission IVV security issues, with 95 and 92, respectively. Therefore, enforcing secure coding practices and verification and validation focused on coding errors would be cost effective ways to improve mission's security. (Flight mission Developers issues dataset did not contain data in the Issue Category.)- In both the Ground and Flight mission IVV issues datasets, the majority of security issues (i.e., 91 and 85, respectively) were introduced in the Implementation phase. In most cases, the phase in which the issues were found was the same as the phase in which they were introduced. The most security related issues of the Flight mission Developers issues dataset were found during Code Implementation, Build Integration, and Build Verification; the data on the phase in which these issues were introduced were not available for this dataset.- The location of security related issues, as the location of software issues in general, followed the Pareto principle. Specifically, for all three datasets, from 86 to 88 the security related issues were located in two to four subsystems.- The severity levels of most security issues were moderate, in all three datasets.- Out of 21 primary security classes, five dominated: Exception Management, Memory Access, Other, Risky Values, and Unused Entities. Together, these classes contributed from around 80 to 90 of all security issues in each dataset. This again proves the Pareto principle of uneven distribution of security issues, in this case across CWE classes, and supports the fact that addressing these dominant security classes provides the most cost efficient way to improve missions' security. The findings presented in this report uncovered the security vulnerability profiles and identified the common trends and dominant classes of security issues, which in turn can be used to select the most efficient secure design and coding best practices compiled by the part of the SARP project team associated with the NASA's Johnson Space Center. In addition, these findings provide valuable input to the NASA IVV initiative aimed at identification of the two 25 CWEs of ground and flight missions.

vulnerability↗
DOE OSTI2017

On the impact of large angle CMB polarization data on cosmological parameters

We study the impact of the large-angle CMB polarization datasets publicly released by the WMAP and Planck satellites on the estimation of cosmological parameters of the ΛCDM model. To complement large-angle polarization, we consider the high resolution (or "high-ℓ") CMB datasets from either WMAP or Planck as well as CMB lensing as traced by Planck's measured four point correlation function. In the case of WMAP, we compute the large-angle polarization likelihood starting over from low resolution frequency maps and their covariance matrices, and perform our own foreground mitigation technique, which includes as a possible alternative Planck 353 GHz data to trace polarized dust. We find that the latter choice induces a downward shift in the optical depth τ, roughly of order 2σ, robust to the choice of the complementary high resolution dataset. When the Planck 353 GHz is consistently used to minimize polarized dust emission, WMAP and Planck 70 GHz large-angle polarization data are in remarkable agreement: by combining them we find τ = 0.066 +0.012-0.013, again very stable against the particular choice for high-ℓ data. We find that the amplitude of primordial fluctuations As, notoriously degenerate with τ, is the parameter second most affected by the assumptions on polarized dust removal, but the other parameters are also affected, typically between 0.5 and 1σ. In particular, cleaning dust with Planck's 353 GHz data imposes a 1σ downward shift in the value of the Hubble constant H0, significantly contributing to the tension reported between CMB based and direct measurements of the present expansion rate. On the other hand, we find that the appearance of the so-called low ℓ anomaly, a well-known tension between the high- and low-resolution CMB anisotropy amplitude, is not significantly affected by the details of large-angle polarization, or by the particular high-ℓ dataset employed.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Deep learning with mixup augmentation for improved pore detection during additive manufacturing

In additive manufacturing (AM), process defects such as keyhole pores are difficult to anticipate, affecting the quality and integrity of the AM-produced materials. Hence, considerable efforts have aimed to predict these process defects by training machine learning (ML) models using passive measurements such as acoustic emissions. This work considered a dataset in which keyhole pores of a laser powder bed fusion (LPBF) experiment were identified using X-ray radiography and then registered both in space and time to acoustic measurements recorded during the LPBF experiment. Due to AM’s intrinsic process controls, where a pore-forming event is relatively rare, the acoustic datasets collected during monitoring include more non-pores than pores. In other words, the dataset for ML model development is imbalanced. Moreover, this imbalanced and sparse data phenomenon remains ubiquitous across many AM monitoring schemes since training data is nontrivial to collect. Hence, we propose a machine learning approach to improve this dataset imbalance and enhance the prediction accuracy of pore-labeled data. Specifically, we investigate how data augmentation helps predict pores and non-pores better. This imbalance is improved using recent advances in data augmentation called Mixup, a weak-supervised learning method. Convolutional neural networks (CNNs) are trained on original and augmented datasets, and an appreciable increase in performance is reported when testing on five different experimental trials. When ML models are trained on original and augmented datasets, they achieve an accuracy of 95% and 99% on test datasets, respectively. We also provide information on how dataset size affects model performance. Lastly, we investigate the optimal Mixup parameters for augmentation in the context of CNN performance.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Multi-fidelity learning for interatomic potentials: low-level forces and high-level energies are all you need

The promise of machine learning interatomic potentials (MLIPs) has led to an abundance of public quantum mechanical (QM) training datasets. The quality of an MLIP is directly limited by the accuracy of the energies and atomic forces in the training dataset. Unfortunately, most of these datasets are computed with relatively low-accuracy QM methods, e.g. density functional theory with a moderate basis set. Due to the increased computational cost of more accurate QM methods, e.g. coupled-cluster theory with a complete basis set (CBS) extrapolation, most high-accuracy datasets are much smaller and often do not contain atomic forces. The lack of high-accuracy atomic forces is quite troubling, as training with force data greatly improves the stability and quality of the MLIP compared to training to energy alone. Because most datasets are computed with a unique level of theory, traditional single-fidelity (SF) learning is not capable of leveraging the vast amounts of published QM data. In this study, we apply multi-fidelity learning (MFL) to train an MLIP to multiple QM datasets of different levels of accuracy, i.e. levels of fidelity. Specifically, we perform three test cases to demonstrate that MFL with both low-level forces and high-level energies yields an extremely accurate MLIP—far more accurate than a SF MLIP trained solely to high-level energies and almost as accurate as a SF MLIP trained directly to high-level energies and forces. Therefore, MFL greatly alleviates the need for generating large and expensive datasets containing high-accuracy atomic forces and allows for more effective training to existing high-accuracy energy-only datasets. Indeed, low-accuracy atomic forces and high-accuracy energies are all that are needed to achieve a high-accuracy MLIP with MFL.

36 MATERIALS SCIENCE↗
DOE OSTI2019

ACTINN: automated identification of cell types in single cell RNA sequencing

Abstract Motivation Cell type identification is one of the major goals in single cell RNA sequencing (scRNA-seq). Current methods for assigning cell types typically involve the use of unsupervised clustering, the identification of signature genes in each cluster, followed by a manual lookup of these genes in the literature and databases to assign cell types. However, there are several limitations associated with these approaches, such as unwanted sources of variation that influence clustering and a lack of canonical markers for certain cell types. Here, we present ACTINN (Automated Cell Type Identification using Neural Networks), which employs a neural network with three hidden layers, trains on datasets with predefined cell types and predicts cell types for other datasets based on the trained parameters. Results We trained the neural network on a mouse cell type atlas (Tabula Muris Atlas) and a human immune cell dataset, and used it to predict cell types for mouse leukocytes, human PBMCs and human T cell sub types. The results showed that our neural network is fast and accurate, and should therefore be a useful tool to complement existing scRNA-seq pipelines. Availability and implementation The codes and datasets are available at https://figshare.com/articles/ACTINN/8967116. Tutorial is available at https://github.com/mafeiyang/ACTINN. All codes are implemented in python. Supplementary information Supplementary data are available at Bioinformatics online.

Ma, Feiyang (ORCID:0000000262600787)↗
DOE OSTI2018

Underwater Fish Detection using Deep Learning for Water Power Applications

Clean energy from oceans and rivers is becoming a reality with the development of new technologies like tidal and instream turbines that generate electricity from naturally flowing water. These new technologies are being monitored for effects on fish and other wildlife using underwater video. Methods for automated analysis of underwater video are needed to lower the costs of analysis and improve accuracy. We have trained a deep learning model, YOLO, to recognize fish in underwater video using three very different datasets recorded at real-world water power sites. Training and testing with examples from all three datasets resulted in a mean average precision (MAP) score of 0.5392. To test how well a model could generalize to new datasets, the model was trained using examples from only two of the datasets and then tested on examples from all three datasets. The resulting model could not recognize fish in the dataset that was not part of the training set. The MAP scores on the other two datasets that were included in the training set were higher than the scores achieved by the model trained on all three datasets. These results indicate that different methods are needed in order to produce a trained model that can generalize to new data sets such as those encountered in real world applications.

fish detection, underwater video, marine and hydro↗
DOE OSTI2025

HPC Campaign Management: Remote data access with user-defined error bound using ADIOS and ZFP

Remote access to large-scale scientific datasets, like those generated by combustion simulations or other high-performance computing (HPC) applications, presents a significant challenge. Downloading entire datasets is often impractical due to their size and the bandwidth limitations of typical networks. To address this challenge, we propose a novel approach that enables efficient remote access to large datasets distributed across multiple facilities. Our method enables technologies to download only the data values of a select variable, in a select region of interest, to a user-defined accuracy. For this purpose, we extended the ADIOS IO library to provide read functions with user-defined accuracy, a remote data server that understands multidimensional selections of specific variables, steps and accuracy from an ADIOS dataset, and which uses lossy compression on the remote site to reduce the data to be transferred back to the client. In addition, our extension of the ADIOS library collects metadata from multiple datasets in small files called Campaign Archives, which can be shared among project participants on any HPC, cloud or laptop, and which can easily facilitate the discovery of content and pointers to the data location as well as remote access to the data by local tools as if data was local. This feature called Campaign Management, enables a group of scientists to manage related datasets stored in multiple files, across multiple facilities as if it was in a single file/database. We demonstrate the effectiveness of our approach using a 1.5 TB dataset from the S3D combustion simulation on Frontier at the Oak Ridge Leadership Facility. Even a single variable from this dataset, at 64 GB, is too large to be processed on a standard laptop. We show two different reading patterns for 2D plots and 3D visualization, with careful settings that a scientist studying combustion data would do and show that running the same Python scripts on Frontier directly takes comparable time than running them on the local laptop with remote access to the data on Frontier.

Podhorszki, Norbert [ORNL] (ORCID:000000019647542X↗