Search NASA⌕ Search

SEARCH · Search NASA

Results for “Selective classification”

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 541 records · Page 30

A SPECTROSCOPIC CATALOG OF THE BRIGHTEST (J < 9) M DWARFS IN THE NORTHERN SKY{sup ,}

We present a spectroscopic catalog of the 1564 brightest (J < 9) M dwarf candidates in the northern sky, as selected from the SUPERBLINK proper motion catalog. Observations confirm 1408 of the candidates to be late-K and M dwarfs with spectral subtypes K7-M6. From the low ({mu} > 40 mas yr{sup -1}) proper motion limit and high level of completeness of the SUPERBLINK catalog in that magnitude range, we estimate that our spectroscopic census most likely includes >90% of all existing, northern-sky M dwarfs with apparent magnitude J < 9. Only 682 stars in our sample are listed in the Third Catalog of Nearby Stars (CNS3); most others are relative unknowns and have spectroscopic data presented here for the first time. Spectral subtypes are assigned based on spectral index measurements of CaH and TiO molecular bands; a comparison of spectra from the same stars obtained at different observatories, however, reveals that spectral band index measurements are dependent on spectral resolution, spectrophotometric calibration, and other instrumental factors. As a result, we find that a consistent classification scheme requires that spectral indices be calibrated and corrected for each observatory/instrument used. After systematic corrections and a recalibration of the subtype-index relationships for the CaH2, CaH3, TiO5, and TiO6 spectral indices, we find that we can consistently and reliably classify all our stars to a half-subtype precision. The use of corrected spectral indices further requires us to recalibrate the {zeta} parameter, a metallicity indicator based on the ratio of TiO and CaH optical bandheads. However, we find that our {zeta} values are not sensitive enough to diagnose metallicity variations in dwarfs of subtypes M2 and earlier ({+-}0.5 dex accuracy) and are only marginally useful at later M3-M5 subtypes ({+-}0.2 dex accuracy). Fits of our spectra to the Phoenix atmospheric model grid are used to estimate effective temperatures. These suggest the existence of a plateau in the M1-M3 subtype range, in agreement with model fits of infrared spectra but at odds with photometric determinations of T{sub eff}. Existing geometric parallax measurements are extracted from the literature for 624 stars, and are used to determine spectroscopic and photometric distances for all the other stars. Active dwarfs are identified from measurements of H{alpha} equivalent widths, and we find a strong correlation between H{alpha} emission in M dwarfs and detected X-ray emission from ROSAT and/or a large UV excess in the GALEX point source catalog. We combine proper motion data and photometric distances to evaluate the (U, V, W) distribution in velocity space, which is found to correlate tightly with the velocity distribution of G dwarfs in the solar neighborhood. However, active stars show a smaller dispersion in their space velocities, which is consistent with those stars being younger on average. Our catalog will be most useful to guide the selection of the best M dwarf targets for exoplanet searches, in particular those using high-precision radial velocity measurements.

79 ASTRONOMY AND ASTROPHYSICS↗

GMFOLD: Subgraph matching for high-throughput DNA-aptamer secondary structure classification and machine learning interpretability

Aptamers are oligonucleotide receptors that bind to their targets with high affinity. Here, we consider aptamers comprised of single-stranded DNA that undergo target-binding-induced conformational changes, giving rise to unique secondary and tertiary structures. Given a specific aptamer primary sequence, there are well-established computational tools (notably mfold) to predict the secondary structure via free energy minimization algorithms. While mfold generates secondary structures for individual sequences, there is a need for a high-throughput process whereby thousands of DNA structures can be predicted in real-time for use in an interactive setting, when combined with aptamer selections that generate candidate pools that are too large to be experimentally interrogated. We developed a new Python code for high-throughput aptamer secondary structure determination (GMfold). GMfold uses subgraph matching methods to group aptamer candidates by secondary structure similarities. We also improve an open-source code, SeqFold, to incorporate subgraph matching concepts. We represent each secondary structure as a lowest-energy bipartite subgraph matching of the DNA graph to itself. These new tools enable thousands of DNA sequences to be compared based on their secondary structures, using machine-learning algorithms. This process is advantageous when analyzing sequences that arise from aptamer selections via systematic evolution of ligands by exponential enrichment (SELEX). This work is a building block for future machine-learning-informed DNA-aptamer selection processes to identify aptamers with improved target affinity and selectivity and advance aptamer biosensors and therapeutics.

Aptamer↗

GOODS-HERSCHEL: IMPACT OF ACTIVE GALACTIC NUCLEI AND STAR FORMATION ACTIVITY ON INFRARED SPECTRAL ENERGY DISTRIBUTIONS AT HIGH REDSHIFT

We explore the effects of active galactic nuclei (AGNs) and star formation activity on the infrared (0.3-1000 {mu}m) spectral energy distributions (SEDs) of luminous infrared galaxies from z = 0.5 to 4.0. We have compiled a large sample of 151 galaxies selected at 24 {mu}m (S {sub 24} {approx}> 100 {mu}Jy) in the GOODS-N and ECDFS fields for which we have deep Spitzer IRS spectroscopy, allowing us to decompose the mid-IR spectrum into contributions from star formation and AGN activity. A significant portion ({approx}25%) of our sample is dominated by an AGN (>50% of the mid-IR luminosity) in the mid-IR. Based on the mid-IR classification, we divide our full sample into four sub-samples: z {approx} 1 star-forming (SF) sources, z {approx} 2 SF sources, AGNs with clear 9.7 {mu}m silicate absorption, and AGNs with featureless mid-IR spectra. From our large spectroscopic sample and wealth of multi-wavelength data, including deep Herschel imaging at 100, 160, 250, 350, and 500 {mu}m, we use 95 galaxies with complete spectral coverage to create a composite SED for each sub-sample. We then fit a two-temperature component modified blackbody to the SEDs. We find that the IR SEDs have similar cold dust temperatures, regardless of the mid-IR power source, but display a marked difference in the warmer dust temperatures. We calculate the average effective temperature of the dust in each sub-sample and find a significant ({approx}20 K) difference between the SF and AGN systems. We compare our composite SEDs to local templates and find that local templates do not accurately reproduce the mid-IR features and dust temperatures of our high-redshift systems. High-redshift IR luminous galaxies contain significantly more cool dust than their local counterparts. We find that a full suite of photometry spanning the IR peak is necessary to accurately account for the dominant dust temperature components in high-redshift IR luminous galaxies.

79 ASTRONOMY AND ASTROPHYSICS↗

Phylogenetic Analysis of Shewanella Strains by DNA Relatedness Derived from Whole Genome Microarray DNA-DNA Hybridization and Comparison with Other Methods

Phylogenetic analyses were done for the Shewanella strains isolated from Baltic Sea (38 strains), US DOE Hanford Uranium bioremediation site [Hanford Reach of the Columbia River (HRCR), 11 strains], Pacific Ocean and Hawaiian sediments (8 strains), and strains from other resources (16 strains) with three out group strains, Rhodopseudomonas palustris, Clostridium cellulolyticum, and Thermoanaerobacter ethanolicus X514, using DNA relatedness derived from WCGA-based DNA-DNA hybridizations, sequence similarities of 16S rRNA gene and gyrB gene, and sequence similarities of 6 loci of Shewanella genome selected from a shared gene list of the Shewanella strains with whole genome sequenced based on the average nucleotide identity of them (ANI). The phylogenetic trees based on 16S rRNA and gyrB gene sequences, and DNA relatedness derived from WCGA hybridizations of the tested Shewanella strains share exactly the same sub-clusters with very few exceptions, in which the strains were basically grouped by species. However, the phylogenetic analysis based on DNA relatedness derived from WCGA hybridizations dramatically increased the differentiation resolution at species and strains level within Shewanella genus. When the tree based on DNA relatedness derived from WCGA hybridizations was compared to the tree based on the combined sequences of the selected functional genes (6 loci), we found that the resolutions of both methods are similar, but the clustering of the tree based on DNA relatedness derived from WMGA hybridizations was clearer. These results indicate that WCGA-based DNA-DNA hybridization is an idea alternative of conventional DNA-DNA hybridization methods and it is superior to the phylogenetics methods based on sequence similarities of single genes. Detailed analysis is being performed for the re-classification of the strains examined.

54 ENVIRONMENTAL SCIENCES↗

Soil Metabolomics Predict Microbial Taxa as Biomarkers of Moisture Status in Soils from a Tidal Wetland

We present observations from a laboratory-controlled study on the impacts of extreme wetting and drying on a wetland soil microbiome. Our approach was to experimentally challenge the soil microbiome to understand impacts on anaerobic carbon cycling processes as the system transitions from dryness to saturation and vice-versa. Specifically, we tested for impacts on stress responses related to shifts from wet to drought conditions. We used a combination of high-resolution data for small organic chemical compounds (metabolites) and biological (community structure based on 16S rRNA gene sequencing) features. Using a robust correlation-independent data approach, we further tested the predictive power of soil metabolites for the presence or absence of taxa. Here, we demonstrate that taking an untargeted, multidimensional data approach to the interpretation of metabolomics has the potential to indicate the causative pathways selecting for the observed bacterial community structure in soils.

16S rRNA gene↗

THE MAGELLANIC QUASARS SURVEY. II. CONFIRMATION OF 144 NEW ACTIVE GALACTIC NUCLEI BEHIND THE SOUTHERN EDGE OF THE LARGE MAGELLANIC CLOUD

We quadruple the number of quasars known behind the Large Magellanic Cloud (LMC) from 56 (42 in the Optical Gravitational Lensing Experiment (OGLE)-III LMC fields) to 200 by spectroscopically confirming 169 (144 new) quasars from a sample of 845 observed candidates in four {approx}3 deg{sup 2} Anglo-Australian Telescope/AAOmega fields south of the LMC center. The candidates were selected based on their Spitzer mid-infrared colors, X-ray emission, and/or optical variability properties in the database of the OGLE microlensing survey. The contaminating sources can be divided into 115 young stellar objects (YSOs), 17 planetary nebulae (PNe), 39 Be and 24 blue stars, 68 red stars, and 12 objects classed as either YSO/PN or blue star/YSO. There are also 402 targets with either featureless spectra or too low signal-to-noise ratios for source classification. Our quasar sample is 50% (30%) complete at I = 18.6 mag (19.3 mag). The newly discovered active galactic nuclei (AGNs) provide many additional reference points for proper motion studies of the LMC, and the sample includes 10 bright AGNs (I < 18 mag) that are potentially suitable for absorption line studies. Their primary use, however, is for detailed studies of quasar variability, as they all have long-term, high cadence, continuously growing light curves from the microlensing surveys of the LMC. Completing the existing Magellanic Quasars Survey fields in the LMC and Small Magellanic Cloud should yield a sample of {approx}700 well-monitored AGNs, and expanding it to the larger regions covered by the OGLE-IV survey should yield a sample of {approx}3600 AGNs.

79 ASTRONOMY AND ASTROPHYSICS↗

PCAfold 2.0—Novel tools and algorithms for low-dimensional manifold assessment and optimization

We describe an update to our open-source Python package, PCAfold, designed to help researchers generate, analyze and improve low-dimensional data manifolds. In the current version, PCAfold 2.0, we introduce novel tools and algorithms for assessing and optimizing low-dimensional manifolds. This includes a method that generates a “map” of local feature sizes that can help pinpoint researchers to problematic regions on a manifold. We introduce a novel cost function that characterizes the quality of a manifold topology with a single number. We develop two algorithms for feature selection based on principal component analysis (PCA) that use the cost function as an objective function to minimize. We introduce a quantity of interest (QoI)-aware dimensionality reduction strategy where data projections are computed using an artificial neural network and are directly optimized towards representing various projection-independent and projection-dependent QoIs. We also introduce an implementation of partition of unity networks (POUnets) for efficient reconstruction of QoIs from low-dimensional manifolds based on combining neural network classification with localized polynomial regression. Our software can be broadly applicable in all domains of science and engineering that aim to reduce data dimensionality, as well as in the fundamental research on representation learning.

97 MATHEMATICS AND COMPUTING↗

NARROW-LINE X-RAY-SELECTED GALAXIES IN THE CHANDRA -COSMOS FIELD. I. OPTICAL SPECTROSCOPIC CATALOG

The COSMOS survey is a large and deep survey with multiwavelength observations of sources from X-rays to the UV, allowing an extensive study of their properties. The central 0.9 deg{sup 2} of the COSMOS field have been observed by Chandra with a sensitivity up to 1.9 × 10{sup −16} erg cm{sup −2} s{sup −1} in the full (0.5–10 keV) band. Photometric and spectroscopic identification of the Chandra -COSMOS (C-COSMOS) sources is available from several catalogs and campaigns. Despite the fact that the C-COSMOS galaxies have a reliable spectroscopic redshift in addition to a spectroscopic classification, the emission-line properties of this sample have not yet been measured. We present here the creation of an emission-line catalog of 453 narrow-line sources from the C-COSMOS spectroscopic sample. We have performed spectral fitting for the more common lines in galaxies ([O ii] λ 3727, [Ne iii] λ 3869, H β , [O iii] λλ 4959, 5007, H α , and [N ii] λλ 6548, 6584). These data provide an optical classification for 151 (i.e., 33%) of the C-COSMOS narrow-line galaxies based on emission-line diagnostic diagrams.

79 ASTRONOMY AND ASTROPHYSICS↗

Modeling Personalized Email Prioritization: Classification-based and Regression-based Approaches

Email overload, even after spam filtering, presents a serious productivity challenge for busy professionals and executives. One solution is automated prioritization of incoming emails to ensure the most important are read and processed quickly, while others are processed later as/if time permits in declining priority levels. This paper presents a study of machine learning approaches to email prioritization into discrete levels, comparing ordinal regression versus classier cascades. Given the ordinal nature of discrete email priority levels, SVM ordinal regression would be expected to perform well, but surprisingly a cascade of SVM classifiers significantly outperforms ordinal regression for email prioritization. In contrast, SVM regression performs well -- better than classifiers -- on selected UCI data sets. This unexpected performance inversion is analyzed and results are presented, providing core functionality for email prioritization systems.

97 MATHEMATICS AND COMPUTING↗

DNABERT-S: pioneering species differentiation with species-aware DNA embeddings

SUMMARY: We introduce DNABERT-S, a tailored genome model that develops species-aware embeddings to naturally cluster and segregate DNA sequences of different species in the embedding space. Differentiating species from genomic sequences (i.e. DNA and RNA) is vital yet challenging, since many real-world species remain uncharacterized, lacking known genomes for reference. Embedding-based methods are therefore used to differentiate species in an unsupervised manner. DNABERT-S builds upon a pre-trained genome foundation model named DNABERT-2. To encourage effective embeddings to error-prone long-read DNA sequences, we introduce Manifold Instance Mixup (MI-Mix), a contrastive objective that mixes the hidden representations of DNA sequences at randomly selected layers and trains the model to recognize and differentiate these mixed proportions at the output layer. We further enhance it with the proposed Curriculum Contrastive Learning (C2LR) strategy. Empirical results on 28 diverse datasets show DNABERT-S's effectiveness, especially in realistic label-scarce scenarios. For example, it identifies twice more species from a mixture of unlabeled genomic sequences, doubles the Adjusted Rand Index (ARI) in species clustering, and outperforms the top baseline's performance in 10-shot species classification with just a 2-shot training. AVAILABILITY AND IMPLEMENTATION: Model, codes, and data are publically available at https://github.com/MAGICS-LAB/DNABERT_S.

Zhou, Zhihan↗

SU-E-J-219: A Dixon Based Pseudo-CT Generation Method for MR-Only Radiotherapy Treatment Planning of the Pelvis and Head and Neck

Purpose: To develop an image processing method for MRI-based generation of electron density maps, known as pseudo-CT (pCT), without usage of model- or atlas-based segmentation, and to evaluate the method in the pelvic and head-neck region against CT. Methods: CT and MRI scans were obtained from the pelvic region of four patients in supine position using a flat table top only for CT. Stratified CT maps were generated by classifying each voxel based on HU ranges into one of four classes: air, adipose tissue, soft tissue or bone.A hierarchical region-selective algorithm, based on automatic thresholding and clustering, was used to classify tissues from MR Dixon reconstructed fat, In-Phase (IP) and Opposed-Phase (OP) images. First, a body mask was obtained by thresholding the IP image. Subsequently, an automatic threshold on the Dixon fat image differentiated soft and adipose tissue. K-means clustering on IP and OP images resulted in a mask that, via a connected neighborhood analysis, allowing the user to select the components corresponding to bone structures.The pCT was estimated through assignment of bulk HU to the tissue classes. Bone-only Digital Reconstructed Radiographs (DRR) were generated as well. The pCT images were rigidly registered to the stratified CT to allow a volumetric and voxelwise comparison. Moreover, pCTs were also calculated within the head-neck region in two volunteers using the same pipeline. Results: The volumetric comparison resulted in differences <1% for each tissue class. A voxelwise comparison showed a good classification, ranging from 64% to 98%. The primary misclassified classes were adipose/soft tissue and bone/soft tissue. As the patients have been imaged on different table tops, part of the misclassification error can be explained by misregistration. Conclusion: The proposed approach does not rely on an anatomy model providing the flexibility to successfully generate the pCT in two different body sites. This research is founded by ZonMw IMDI Programme, project name: “RASOR sharp: MRI based radiotherapy planning using a single MRI sequence”, project number: 10-104003010.

60 APPLIED LIFE SCIENCES↗

Data object classification using feature generation through crowdsourcing

In a computing device that implements a data object classification tool, a method for classifying data may include detecting change in spatial coordinates for each of at least two of a set of data objects within a canvas space. Each of the data objects may be associated with a vector of features. A rule set may be generated based on the vector of features associated with each of the at least two data objects. The rule set may use feature(s) that explain the changed spatial coordinates. The data objects may be selectively rearranged within the canvas space by applying the generated rule set to any remaining data objects among the set of data objects so as to assign spatial coordinates to the remaining objects. For each of the data objects, the spatial coordinates may be stored as new semantic feature(s) within the vector of features for that data object.

97 MATHEMATICS AND COMPUTING↗

PHOTOMETRIC REDSHIFTS FOR QUASARS IN MULTI-BAND SURVEYS

The Multi Layer Perceptron with Quasi Newton Algorithm (MLPQNA) is a machine learning method that can be used to cope with regression and classification problems on complex and massive data sets. In this paper, we give a formal description of the method and present the results of its application to the evaluation of photometric redshifts for quasars. The data set used for the experiment was obtained by merging four different surveys (Sloan Digital Sky Survey, GALEX, UKIDSS, and WISE), thus covering a wide range of wavelengths from the UV to the mid-infrared. The method is able (1) to achieve a very high accuracy, (2) to drastically reduce the number of outliers and catastrophic objects, and (3) to discriminate among parameters (or features) on the basis of their significance, so that the number of features used for training and analysis can be optimized in order to reduce both the computational demands and the effects of degeneracy. The best experiment, which makes use of a selected combination of parameters drawn from the four surveys, leads, in terms of {Delta}z{sub norm} (i.e., (z{sub spec} - z{sub phot})/(1 + z{sub spec})), to an average of {Delta}z{sub norm} = 0.004, a standard deviation of {sigma} = 0.069, and a median absolute deviation, MAD = 0.02, over the whole redshift range (i.e., z{sub spec} {<=} 3.6), defined by the four-survey cross-matched spectroscopic sample. The fraction of catastrophic outliers, i.e., of objects with photo-z deviating more than 2{sigma} from the spectroscopic value, is <3%, leading to {sigma} = 0.035 after their removal, over the same redshift range. The method is made available to the community through the DAMEWARE Web application.

79 ASTRONOMY AND ASTROPHYSICS↗

Oxidative DNA damage of peripheral blood polymorphonuclear leukocytes, selectively induced by chronic arsenic exposure, is associated with extent of arsenic-related skin lesions

There is increasing evidence that oxidative stress is an important risk factor for arsenic-related diseases. Peripheral blood leukocytes constitute an important defense against microorganisms or pathogens, while the research on the impact of chronic arsenic exposure on peripheral blood leukocytes is much more limited, especially at low level arsenic exposure. The purpose of the present study was to explore whether chronic arsenic exposure affects oxidative stress of peripheral blood leukocytes and possible linkages between oxidative stress and arsenic-induced skin lesions. 75 male inhabitants recruited from an As-endemic region of China were investigated in the present study. The classification of arsenicosis was based on the degree of skin lesions. Arsenic levels were measured in drinking water and urine by Atomic Fluorescence Spectroscopy. Urinary 8-hydroxy-2′-deoxyguanosine (8-OHdG) was tested by Enzyme-Linked Immunosorbent Assay. 8-OHdG of peripheral blood leukocytes was evaluated using immunocytochemical staining. 8-OHdG-positive reactions were only present in polymorphonuclear leukocytes (PMNs), but not in monocytes (MNs). The 8-OHdG staining of PMN cytoplasm was observed in all investigated populations, while the 8-OHdG staining of PMN nuclei was frequently found along with the elevated amounts of cell debris in individuals with skin lesion. Urinary arsenic levels were increased in the severe skin lesion group compared with the normal group. No relationship was observed between drinking water arsenic or urine 8-OHdG and the degree of skin lesions. These findings indicated that the target and persistent oxidative stress in peripheral blood PMNs may be employed as a sensitive biomarker directly to assess adverse health effects caused by chronic exposure to lower levels of arsenic. -- Highlights: ► Male inhabitants were investigated from an As-endemic region of China. ► 8-OHdG-positive reactions were only present in polymorphonuclear leukocytes (PMNs). ► 8-OHdG staining of PMN nuclei was paralleled by increased debris of cells. ► Oxidative DNA damage of PMNs is associated with arsenic-related skin lesions.

60 APPLIED LIFE SCIENCES↗

Deep graph representations embed network information for robust disease marker identification

We report that the accurate disease diagnosis and prognosis based on omics data rely on the effective identification of robust prognostic and diagnostic markers that reflect the states of the biological processes underlying the disease pathogenesis and progression. In this article, we present GCNCC, a Graph Convolutional Network-based approach for Clustering and Classification, that can identify highly effective and robust network-based disease markers. Based on a geometric deep learning framework, GCNCC learns deep network representations by integrating gene expression data with protein interaction data to identify highly reproducible markers with consistently accurate prediction performance across independent datasets possibly from different platforms. GCNCC identifies these markers by clustering the nodes in the protein interaction network based on latent similarity measures learned by the deep architecture of a graph convolutional network, followed by a supervised feature selection procedure that extracts clusters that are highly predictive of the disease state. By benchmarking GCNCC based on independent datasets from different diseases (psychiatric disorder and cancer) and different platforms (microarray and RNA-seq), we show that GCNCC outperforms other state-of-the-art methods in terms of accuracy and reproducibility.

59 BASIC BIOLOGICAL SCIENCES↗

Botnet behaviour analysis: How would a data analytics-based system with minimum a priori information perform?

Botnets, as one of the most aggressive threats, has used different techniques, topologies, and communication protocols in different stages of their lifecycle since 2003. Hence, identifying botnets has become very challenging specifically given that they can upgrade their methodology at any time. Various detection approaches have been proposed by the cyber-security researchers, focusing on different aspects of these threats. In this work, 5 different botnet detection approaches are investigated. These systems are selected based on the technique used and type of data used where 2 are public rule–based systems (BotHunter and Snort) and the other 3 use machine learning algorithm with different feature extraction methods (packet payload based and traffic flow based). On the other hand, 4 of these systems are based on a priori knowledge while one is using minimum a priori information. The objective here is to evaluate the effectiveness of these approaches under different scenarios (eg, multi-botnet and single-botnet classifications) as well as exploring how a system with minimum a priori information would perform. The goal is to investigate if a system with minimum a priori information could result in a competitive performance compared to systems using a priori knowledge. The evaluation is shown on 24 publicly available botnet data sets. Results indicate that a machine learning–based system with minimum a priori information not only achieves a very high performance but also generalizes much better than the other systems evaluated on a wide range of botnet structures (from centralized to decentralized botnets).

97 MATHEMATICS AND COMPUTING↗

Machine Learning-Based Identification of the Interface Regions for Coupling Local and Nonlocal Models

Local-nonlocal coupling approaches provide a means to combine the computational efficiency of local models and the accuracy of nonlocal models. However, the coupling process can be challenging, requiring expertise to identify the interface between local and nonlocal regions. Here, this study introduces a machine learning-based approach to automatically detect the regions in which the local and nonlocal models should be used. The method uses loading functions evaluated at grid points to decide the model selection at those points. Training of the networks is based on datasets provided by classes of loading functions for which reference coupling configurations are computed using accurate coupled solutions, where accuracy is measured in terms of the relative error between the solution to the coupling approach and the solution to the nonlocal model. We study two approaches that vary in data structure. The first, the full-domain input data approach, uses the entire load vector and outputs a complete label vector, performing a global classification. The second, a window-based approach, processes loads into windows and addresses the problem as a node-wise classification where each window's central point is classified individually. The classification problems are solved via deep learning algorithms based on convolutional neural networks. The performance of these approaches is studied on one-dimensional numerical examples using F1-scores and accuracy metrics. Notably, the windowing approach achieves an accuracy of 0.96 and an F1-score of 0.97, highlighting its potential to automate coupling processes effectively and enhance computational efficiency in material science applications.

97 MATHEMATICS AND COMPUTING↗

How does ion temperature gradient turbulence depend on magnetic geometry? Insights from data and machine learning

Magnetic geometry has a significant effect on the level of turbulent transport in fusion plasmas. Here, we model and analyse this dependence using multiple machine learning methods and a dataset of >200 000 nonlinear gyrokinetic simulations of ion-temperature-gradient turbulence in diverse non-axisymmetric geometries. The dataset is generated using a large collection of both optimised and randomly generated stellarator equilibria. At fixed gradients and other input parameters, the turbulent heat flux varies between geometries by several orders of magnitude. Trends are apparent among the configurations with particularly high or particularly low heat flux. Regression and classification techniques from machine learning are then applied to extract patterns in the dataset. Due to a symmetry of the gyrokinetic equation, the heat flux and regressions thereof should be invariant to translations of the raw features in the parallel coordinate, similar to translation invariance in computer vision applications. Multiple regression models including convolutional neural networks (CNNs) and decision trees can achieve reasonable predictive power for the heat flux in held-out test configurations, with highest accuracy for the CNNs. Using Spearman correlation, sequential feature selection and Shapley values to measure feature importance, it is consistently found that the most important geometric lever on the heat flux is the flux surface compression in regions of bad curvature. The second most important geometric feature relates to the magnitude of geodesic curvature. These two features align remarkably with surrogates that have been proposed based on theory, while the methods here allow a natural extension to more features for increased accuracy. The dataset, released with this publication, may also be used to test other proposed surrogates, and we find that many previously published proxies do correlate well with both the heat flux and stability boundary.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗