Search NASA⌕ Search

SEARCH · Search NASA

Results for “data set”

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 1,297 records · Page 72

Unsteady Pressure-Sensitive-Paint Shot Noise Reduction

In the study of unsteady Pressure-Sensitive Paint (uPSP) image data sets, pixel intensity values collected by Phantom cameras from multiple perspectives are used to calculate the surface pressure of an experimental model. This paper focuses on mitigation of shot noise which is introduced into the image data set by the cameras. Shot noise impacts the quality of data collected, increasing the system error in its pressure value calculations. This paper characterises statistical methods to reduce shot noise within experimental data by taking an average or weighted average across an area of pixels of their respective counts. Each combined-pixel has a higher total effective intensity value and therefore exhibits reduced noise throughout its time history. This is demonstrated by calculating the standard deviations of the combined pixel values through time, which decrease as the combined pixel area increases. Finally, these methods are compared across experimental wind tunnel conditions to examine their effectiveness at different Mach numbers. Ultimately, the results are used to recommend a blur filter type and size which will be applied to the raw image data prior to processing, as well as a minimum camera resolution for future experiments. Downstream of the filter application, the software to convert the 2D image data sets into surface pressure readings on a 3D grid use a ratio of decimation that assigns multiple pixels to each 3D grid node. This decimation ratio will also be aligned with the size of the blur filter, resulting in a more manageable data set size and compromised spatial resolution. Combined with the effect of the blur filter, the resulting pixel intensity readings will exhibit significantly reduced shot noise, reducing the magnitude of uncertainty and error in the final calculated pressure. For future tests, the camera spatial resolution must be sufficient to capture small flow and model features even after the recommended blur filter is applied. The image data set is taken from a uPSP experiment of the Space Launch System Block 1 cargo vehicle model in September 2019.

Lucy Tang↗

Unsteady Pressure-Sensitive Paint (UPSP) Shot Noise Reduction

In the study of unsteady Pressure-Sensitive Paint (uPSP) image data sets, pixel intensity values collected by Phantom cameras from multiple perspectives are used to calculate the surface pressure of an experimental model. This paper focuses on mitigation of shot noise which is introduced into the image data set by the cameras. Shot noise impacts the quality of data collected, increasing the system error in its pressure value calculations. This paper characterises statistical methods to reduce shot noise within experimental data by taking an average or weighted average across an area of pixels of their respective counts. Each combined-pixel has a higher total effective intensity value and therefore exhibits reduced noise throughout its time history. This is demonstrated by calculating the standard deviations of the combined pixel values through time, which decrease as the combined pixel area increases. Finally, these methods are compared across experimental wind tunnel conditions to examine their effectiveness at different Mach numbers. Ultimately, the results are used to recommend a blur filter type and size which will be applied to the raw image data prior to processing, as well as a minimum camera resolution for future experiments. Downstream of the filter application, the software to convert the 2D image data sets into surface pressure readings on a 3D grid use a ratio of decimation that assigns multiple pixels to each 3D grid node. This decimation ratio will also be aligned with the size of the blur filter, resulting in a more manageable data set size and compromised spatial resolution. Combined with the effect of the blur filter, the resulting pixel intensity readings will exhibit significantly reduced shot noise, reducing the magnitude of uncertainty and error in the final calculated pressure. For future tests, the camera spatial resolution must be sufficient to capture small flow and model features even after the recommended blur filter is applied. The image data set is taken from a uPSP experiment of the Space Launch System Block 1 cargo vehicle model in September 2019.

Lucy Tang↗

SoyFACE Fumigation Data Files

This data set is related to the SoyFACE experiments, which are open-air agricultural climate change experiments that have been conducted since 2001. The fumigation experiments take place at the SoyFACE farm and facility in Champaign County, Illinois during the growing season of each year, typically between June and October. - The "SoyFACE Plot Information 2001 to 2021" file contains information about each year of the SoyFACE experiments, including the fumigation treatment type (CO2, O3, or a combination treatment), the crop species, the plots (also referred to as 'rings' and labeled with numbers between 2 and 31) used in each experiment, important experiment dates, and the target concentration levels or 'setpoints' for CO2 and O3 in each experiment. - This data set includes files with minute readings of the fumigation levels ( "SoyFACE 1-Minute Fumigation Data Files" folder) from the SoyFACE experiments. The "Soyface 1-Minute Fumigation Data Files" folder contains sub-folders for each year of the experiments, each of which contains sub-folders for each ring used in that year's experiments. This data set also includes hourly data files for the fumigation experiments ( "SoyFACE Hourly Fumigation Data Files" folder) created from the 1-minute files, and hourly ambient/weather data files for each year of the experiments ( "Hourly Weather and Ambient Data Files" folder). The ambient CO2 and O3 data are collected at SoyFACE, and the weather data are collected from the SURFRAD and WARM weather stations located near the SoyFACE farm. - The "Fumigation Target Percentages" file shows how much of the time the CO2 and O3 fumigation levels are within a 10 or 20 percent margin of the target levels when the fumigation system is turned on. - The "Matlab Files" folder contains custom code (Aspray, E.K.) that was used to clean the "SoyFACE 1-Minute Fumigation Data" files and to generate the "SoyFACE Hourly Fumigation Data" and "Fumigation Target Percentages" files. Code information can be found in the "SoyFACE Hourly Fumigation Data Explanation" file. - Finally, the " * Explanation" files contain information about the column names, units of measurement, and other pertinent information for each data file. * NOTE: We have identified some files in the “SoyFACE 1-Minute Fumigation Data Files” folder in our SoyFACE data set submission that were not downloaded properly - the files were present in the folder, but the actual files were empty. V3 ensures that there are no longer any empty files in the data set.

agricultural↗

Machine learning analysis of RB-TnSeq fitness data predicts functional gene modules in Pseudomonas putida KT2440

ABSTRACT There is growing interest in engineering Pseudomonas putida KT2440 as a microbial chassis for the conversion of renewable and waste-based feedstocks, and metabolic engineering of P. putida relies on the understanding of the functional relationships between genes. In this work, independent component analysis (ICA) was applied to a compendium of existing fitness data from randomly barcoded transposon insertion sequencing (RB-TnSeq) of P. putida KT2440 grown in 179 unique experimental conditions. ICA identified 84 independent groups of genes, which we call fModules (“functional modules”), where gene members displayed shared functional influence in a specific cellular process. This machine learning-based approach both successfully recapitulated previously characterized functional relationships and established hitherto unknown associations between genes. Selected gene members from fModules for hydroxycinnamate metabolism and stress resistance, acetyl coenzyme A assimilation, and nitrogen metabolism were validated with engineered mutants of P. putida . Additionally, functional gene clusters from ICA of RB-TnSeq data sets were compared with regulatory gene clusters from prior ICA of RNAseq data sets to draw connections between gene regulation and function. Because ICA profiles the functional role of several distinct gene networks simultaneously, it can reduce the time required to annotate gene function relative to manual curation of RB-TnSeq data sets. IMPORTANCE This study demonstrates a rapid, automated approach for elucidating functional modules within complex genetic networks. While Pseudomonas putida randomly barcoded transposon insertion sequencing data were used as a proof of concept, this approach is applicable to any organism with existing functional genomics data sets and may serve as a useful tool for many valuable applications, such as guiding metabolic engineering efforts in other microbes or understanding functional relationships between virulence-associated genes in pathogenic microbes. Furthermore, this work demonstrates that comparison of data obtained from independent component analysis of transcriptomics and gene fitness datasets can elucidate regulatory-functional relationships between genes, which may have utility in a variety of applications, such as metabolic modeling, strain engineering, or identification of antimicrobial drug targets.

09 BIOMASS FUELS↗

Smoky Mountain Data Challenge 2020: An Open Call to Solve Data Problems in the Areas of Neutron Science, Material Science, Urban Modeling and Dynamics, Geophysics, and Biomedical Informatics

The 2020 Smoky Mountains Computational Sciences and Engineering Conference enlists research scientists from across Oak Ridge National Laboratory (ORNL) to be data sponsors and help create data analytics challenges for eminent data sets at the laboratory. This work describes the significance of each of the seven data sets and their as- sociated challenge questions. The challenge questions for each data set were required to cover multiple difficulty levels. An international call for participation was sent to students, and researchers asking them to form teams of up to four people to apply novel data analytics techniques to these data sets.

Parete-Koon, Suzanne↗

Coordinated field study for CaPE: Analysis of energy and water budgets

The objectives of this hydrologic cycle study are to understand and model (1) surface energy and land-atmosphere water transfer processes, and (2) interactions between convective storms and surface energy fluxes. A surface energy budget measurement campaign was carried out by an interdisciplinary science team during the period July 8 - August 19, 1991 as part of the Convection and Precipitation/Electrification Experiment (CaPE) in the vicinity of Cape Canaveral, FL. Among the research themes associated with CaPE is the remote estimation of rainfall. Thus, in addition to surface radiation and energy budget measurements, surface mesonet, special radiosonde, precipitation, high-resolution satellite (SPOT) data, geosynchronous (GOES) and polar orbiting (DMSP SSM/I, OLS; NOAA AVHRR) satellite data, and high altitude airplane data (AMPR, MAMS, HIS) were collected. Initial quality control of the seven surface flux station data sets has begun. Ancillary data sets are being collected and assembled for analysis. Browsing of GOES and radar data has begun to classify days as disturbed/undisturbed to identify the larger scale forcing of the pre-convective environment, convection storms and precipitation. The science analysis plan has been finalized and tasks assigned to various investigators.

Goodman, Steven J.↗

Improved joint X-ray and neutron refinement procedure in Phenix

Neutron diffraction is one of the three crystallographic techniques (X-ray, neutron and electron diffraction) used to determine the atomic structures of molecules. Its particular strengths derive from the fact that H (and D) atoms are strong neutron scatterers, meaning that their positions, and thus protonation states, can be derived from crystallographic maps. However, because of technical limitations and experimental obstacles, the quality of neutron diffraction data is typically much poorer (completeness, resolution and signal to noise) than that of X-ray diffraction data for the same sample. Further, refinement is more complex as it usually requires additional parameters to describe the H (and D) atoms. The increase in the number of parameters may be mitigated by using the `riding hydrogen' refinement strategy, in which the positions of H atoms without a rotational degree of freedom are inferred from their neighboring heavy atoms. However, this does not address the issues related to poor data quality. Therefore, neutron structure determination often relies on the presence of an X-ray data set for joint X-ray and neutron (XN) refinement. In this approach, the X-ray data serve to compensate for the deficiencies of the neutron diffraction data by refining one model simultaneously against the X-ray and neutron data sets. To be applicable, it is assumed that both data sets are highly isomorphous, and preferably collected from the same crystals and at the same temperature. However, the approach has a number of limitations that are discussed in this work by comparing four separately re-refined neutron models. To address the limitations, a new method for joint XN refinement is introduced that optimizes two different models against the different data sets. This approach is tested using neutron models and data deposited in the Protein Data Bank. The efficacy of refining models with H atoms as riding or as individual atoms is also investigated.

36 MATERIALS SCIENCE↗

First Discoveries of z > 6 Quasars with the DECam Legacy Survey and UKIRT Hemisphere Survey

In this paper, we present the first discoveries from a survey of z ≳ 6 quasars using imaging data from the DECam Legacy Survey (DECaLS) in the optical, the UKIRT Deep Infrared Sky Survey (UKIDSS) and a preliminary version of the UKIRT Hemisphere Survey (UHS) in the near-IR, and ALLWISE in the mid-IR. DECaLS will image 9000 deg 2 of sky down to z AB ~ 23.0, and UKIDSS and UHS will map the northern sky at 0 < decl. < +60°, reaching J VEGA ~ 19.6 (5-σ). The combination of these data sets allows us to discover quasars at redshift z ≳ 7 and to conduct a complete census of the faint quasar population at z ≳ 6. In this paper, we report on the selection method of our search, and on the initial discoveries of two new, faint z ≳ 6 quasars and one new z = 6.63 quasar in our pilot spectroscopic observations. The two new z ~ 6 quasars are at z = 6.07 and z = 6.17 with absolute magnitudes at rest-frame wavelength 1450 Å being M 1450 = -25.83 and M 1450 = -25.76, respectively. These discoveries suggest that we can find quasars close to or fainter than the break magnitude of the Quasar Luminosity Function (QLF) at z ≳ 6. The new z = 6.63 quasar has an absolute magnitude of M 1450 = -25.95. This demonstrates the potential of using the combined DECaLS and UKIDSS/UHS data sets to find z ≳ 7 quasars. Finally, extrapolating from previous QLF measurements, we predict that these combined data sets will yield ~200 z ~ 6 quasars to z AB < 21.5, ~1000 z ~ 6 quasars to z AB < 23, and ~30 quasars at z > 6.5 to J VEGA < 19.5.

79 ASTRONOMY AND ASTROPHYSICS↗

Measurements of the thermal and ionization state of the intergalactic medium during the cosmic afternoon

We perform the first measurement of the thermal and ionization state of the intergalactic medium (IGM) across 0.9 < z < 1.5 using 301 Ly α absorption lines fitted from 12 archival Hubble Space Telescope Space Telescope Imaging Spectrograph quasar spectra. We employ the machine-learning-based inference method that uses joint Doppler parameter–column density (⁠b-N HI ⁠) distributions obtained from Ly α forest decomposition. Our results show that the Γ HI photoionization rates, ⁠, agree with recent ultraviolet background synthesis models, with log(Γ HI /s -1 ) = $-11.79^{+0.18}_{-0.15}$, $-11.98^{+0.09}_{-0.09}$⁠, and $-12.32^{+0.10}_{-0.12}$⁠, at z = 1.4, 1.2, and 1, respectively. We obtain the IGM temperature at the mean density, T 0 ⁠, and the adiabatic index, γ⁠, as [log(T 0 /K), γ] = $[4.13^{+0.12}_{-0.10}, 1.34^{+0.10}_{-0.15}]$, $[3.79^{+0.11}_{-0.11}, 1.70^{+0.09}_{-0.09}]$, and $[4.12^{+0.15}_{-0.25}, 1.34^{+0.21}_{-0.26}]$ at z = 1.4⁠, 1.2, and 1. Our measurements of T 0 at z = 1.4 and 1.2 are consistent with the trend predicted from previous z < 3 temperature measurements and theoretical expectations, where the IGM cools down after $He\tiny{II}$ reionization in the absence of any non-standard heating. However, our T 0 measurement at z = 1 unexpectedly high IGM temperature. Given the relatively large uncertainty in these measurements, where σ T$_0$ ~ 5000 K, mostly emanating from the limited size of our data set, we cannot conclude whether the IGM cools down as expected. Lastly, we generate mock data sets to test the constraining power of future measurement with larger data sets. The results demonstrate that, with redshift path-length Δz ~ 2 for each redshift bin, three times the current data set, we can constrain the T 0 of IGM within 1500 K, which would be sufficient to constrain the IGM thermal history at z < 1.5 conclusively.

79 ASTRONOMY AND ASTROPHYSICS↗

SU-E-T-174: Evaluation of the Optimal Intensity Modulated Radiation Therapy Plans Done On the Maximum and Average Intensity Projection CTs

Purpose: To determine the difference in coverage between plans done on average intensity projection and maximum intensity projection CT data sets for lung patients and to establish correlations between different factors influencing the coverage. Methods: For six lung cancer patients, 10 phases of equal duration through the respiratory cycle, the maximum and average intensity projections (MIP and AIP) from their 4DCT datasets were obtained. MIP and AIP datasets had three GTVs delineated (GTVaip — delineated on AIP, GTVmip — delineated on MIP and GTVfus — delineated on each of the 10 phases and summed up). From the each GTV, planning target volumes (PTV) were then created by adding additional margins. For each of the PTVs an IMRT plan was developed on the AIP dataset. The plans were then copied to the MIP data set and were recalculated. Results: The effective depths in AIP cases were significantly smaller than in MIP (p < 0.001). The Pearson correlation coefficient of r = 0.839 indicates strong degree of positive linear relationship between the average percentage difference in effective depths and average PTV coverage on the MIP data set. The V2 0 Gy of involved lung depends on the PTV coverage. The relationship between PTVaip mean CT number difference and PTVaip coverage on MIP data set gives r = 0.830. When the plans are produced on MIP and copied to AIP, r equals −0.756. Conclusion: The correlation between the AIP and MIP data sets indicates that the selection of the data set for developing the treatment plan affects the final outcome (cases with high average percentage difference in effective depths between AIP and MIP should be calculated on AIP). The percentage of the lung volume receiving higher dose depends on how well PTV is covered, regardless of on which set plan is done.

60 APPLIED LIFE SCIENCES↗

The Rise of Neural Networks for Materials and Chemical Dynamics

Machine learning (ML) is quickly becoming a premier tool for modeling chemical processes and materials. ML-based force fields, trained on large data sets of high-quality electron structure calculations, are particularly attractive due their unique combination of computational efficiency and physical accuracy. This Perspective summarizes some recent advances in the development of neural network-based interatomic potentials. Designing high-quality training data sets is crucial to overall model accuracy. One strategy is active learning, in which new data are automatically collected for atomic configurations that produce large ML uncertainties. Another strategy is to use the highest levels of quantum theory possible. Transfer learning allows training to a data set of mixed fidelity. A model initially trained to a large data set of density functional theory calculations can be significantly improved by retraining to a relatively small data set of expensive coupled cluster theory calculations. These advances are exemplified by applications to molecules and materials.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Cost of Fish Exclusion and Passage Technologies for Hydropower

Hydropower represents a reliable source of renewable energy and accounts for approximately 7% of the total electrical generation in the United States. Future expansion of hydropower is likely to be in the form of either smaller new stream development projects or powering existing non-powered dams. For these new projects to be successful, careful analysis of risks, costs, and uncertainty to offset reduced power production as well as ensuring the protection and safe passage of migratory fish to gain public support, will be required. Exclusion and passage are two common approaches to protect fish from entrainment and impingement at hydropower facilities. The thresholds for entrainment risk and requirements for exclusion and passage often differ depending on the species involved, the characteristics of the facility, and the goals of stakeholders. While the costs associated with environmental mitigations represent a large proportion of the total costs required for the licensing of hydropower facilities, little quantitative information is present within the literature regarding the specific costs of fish exclusion and passage. Working with FOA awardee Natel Energy, scientists at Oak Ridge National Laboratory were tasked with assessing the capital construction costs for downstream fish exclusion and passage infrastructure. This report used keyword searches of an existing environmental mitigation cost data set and manual extraction of additional cost data associated with protection, mitigation, and enhancement (PM&E) measures related to positive barrier screening and passage from regulatory licensing documents available in the Federal Energy Regulatory Commission (FERC) eLibrary. This approach yielded a total of 50 PM&E mitigation measures with estimated capital construction costs pertaining to positive barrier screens, 142 pertaining to passage studies, and 26 pertaining to passage-related studies. PM&E measures associated with positive barrier screens represented <10% of the 171 total FERC project dockets available in the data set. These data were highly skewed toward conventional relicensing projects, as <7% were associated with new stream development (NSD) projects. Results from these data indicate highly variable costs associated with fish screening, with flow-normalized costs one to two orders of magnitude higher for screening with the highest exclusion capability (≤0.09 in. spacing) compared with coarser screening (1 to 2 in.). Furthermore, estimated capital costs of passage infrastructure were positively related to the scale of the project based on installed capacity for some, but not all, types of passage. These data provide an initial baseline for estimating exclusion and passage costs for hydropower development and may help developers consider options for more fish-friendly generation technologies, though gaps remain relating to a lack of data, particularly for NSD projects. More data may still be available within the FERC eLibrary, but significant effort will be required to manually identify and extract the data for future analyses.

13 HYDRO ENERGY↗

Hot-Fire Testing of 5N and 22N HPGP Thrusters

This hot-fire test continues NASA investigation of green propellant technologies for future missions. To show the potential for green propellants to replace some hydrazine systems in future spacecraft, NASA Marshall Space Flight Center (MSFC) is continuing to embark on hot-fire test campaigns with various green propellant blends.NASA completed hot-fire testing of 5N and 22N HPGP thrusters at the Marshall Space Flight Center’s Component Development Area altitude test stand in April 2015. Both thrusters are ground test articles and not flight ready units, but are representative of potential flight hardware with a known path towards flight application. The purpose of the 5N testing was to perform facility check-outs and generate a small set of data for comparison to ECAPS and Orbital ATK data sets. The 5N thruster performed as expected with thrust and propellant flow-rate data generated that are similar to previous testing at Orbital ATK. Immediately following the 5N testing, and using the same facility, the 22N testing was conducted on the same test stand with the purpose of demonstrating the 22N performance. The results of 22N testing indicate it performed as expected.The results of the hot-fire testing are presented in this paper and presentation.

Burnside, Christopher G.↗

Satellite Maps and Relevant Compositional Properties of PM2.5 in Difficult Winter Situations and Comparisons to DISCOVER-AQ Airborne Sampling

Mediterranean-climate regions like California's San Joaquin Valley are subject to severe wintertime particulate pollution affecting public health. We present maps of episodes and particulate diagnostics to aid diagnosis and amelioration. For abatement at sources, we require an understanding of sources and transport. Remote sensing should be of aid, but radiance-to-particle relationships are far different from methods which have been of use in the Eastern USA, Northern and Central Europe. Here are the problems: (a) Thin if very polluted mixed layers (MLs) yield optical depths, AOD, near the detection level, (b) bright and quite variegated surfaces (c) Unusual particle composition (e.g., predominance of NH4NO3 and fireplace buning aerosol), which complicate the relationship of AOD to PM2.5. Specialized analysis of MODIS-Aqua data to obtain AOD using the multi-angle (MAIAC) technique employed by Lyapustin and Wang. Meteorological analyses like NOAA's Rapid Analysis Product (RAP, or newer products like HRRR), which are formulated to remain close to observations (e.g. of water), provide useful ML information corroborated by DISCOVER-AQ in-situ and lidar observations. The many PM2.5 measurements allow a calibration of these products and thus maps of aerosol on many successive aerosol buildups. These calibrations also allow insight into compositional information relevant to MODIS retrievals, the product of aerosol density and specific scattering. We have found that the rich in-situ, lidar, and sun-photometer data sets of NASA'S DISCOVER-AQ data set (2013) of great aid. We will highlight the most interesting of many intercomparisons possible with this rich data set. We conclude with a description of new work to extend these insights to similar regions, e.g. the Imperial Valley of California, the Po Valley and maritime Southern Europe, and the litoral regions of Israel.

aerosols↗

Autonomous, Full-Time Cloud Profiling at Arm Sites with Micro Pulse Lidar

Since the early 1990's technology advances permit ground based lidar to operate full time and profile all significant aerosol and cloud structure of the atmosphere up to the limit of signal attenuation. These systems are known as Micro Pulse Lidars (MPL), as referenced by Spinhirne (1993), and were first in operation at DOE Atmospheric Radiation Measurement (ARM) sites. The objective of the ARM program is to improve the predictability of climate change, particularly as it relates to cloud-climate feedback. The fundamental application of the MPL systems is towards the detection of all significant hydrometeor layers, to the limit of signal attenuation. The heating and cooling of the atmosphere are effected by the distribution and characteristics of clouds and aerosol concentration. Aerosol and cloud retrievals in several important areas can only be adequately obtained with active remote sensing by lidar. For cloud cover, the height and related emissivity of thin clouds and the distribution of base height for all clouds are basic parameters for the surface radiation budget, and lidar is essetial for accurate measurements. The ARM MPL observing network represents the first long-term, global lidar study known within the community. MPL systems are now operational at four ARM sites. A six year data set has been obtained at the original Oklahoma site, and there are several years of observations at tropical and artic sites. Observational results include cloud base height distributions and aerosol profiles. These expanding data sets offer a significant new resource for cloud, aerosol and atmospheric radiation analysis. The nature of the data sets, data processing algorithms, derived parameters and application results are presented.

Spinhirne, James D.↗

A modeling analysis program for the JPL table mountain Io sodium cloud

Progress and achievements in the first year are discussed in three main areas: (1) review and assessment of the massive JPL Table Mountain Io sodium cloud data set, (2) formulation and execution of a plan to perform further processing of this data set, and (3) initiation of modeling activities. The complete 1976-79 and 1981 data sets are reviewed. Particular emphasis is placed on the superior 1981 Region B/C images which provide a rich base of information for studying the structure and escape of gases from Io as well as possible east-west and magnetic longitudinal asymmetries in the plasma torus. A data processing plan is developed and is undertaken by the Multimission Image Processing Laboratory of JPL for the purpose of providing a more refined and complete data set for our modeling studies in the second year. Modeling priorities are formulated and initial progress in achieving these goals is reported.

Smyth, W. H.↗

Comparison of Bayesian network and support vector machine models for two-year survival prediction in lung cancer patients treated with radiotherapy

Purpose: Classic statistical and machine learning models such as support vector machines (SVMs) can be used to predict cancer outcome, but often only perform well if all the input variables are known, which is unlikely in the medical domain. Bayesian network (BN) models have a natural ability to reason under uncertainty and might handle missing data better. In this study, the authors hypothesize that a BN model can predict two-year survival in non-small cell lung cancer (NSCLC) patients as accurately as SVM, but will predict survival more accurately when data are missing. Methods: A BN and SVM model were trained on 322 inoperable NSCLC patients treated with radiotherapy from Maastricht and validated in three independent data sets of 35, 47, and 33 patients from Ghent, Leuven, and Toronto. Missing variables occurred in the data set with only 37, 28, and 24 patients having a complete data set. Results: The BN model structure and parameter learning identified gross tumor volume size, performance status, and number of positive lymph nodes on a PET as prognostic factors for two-year survival. When validated in the full validation set of Ghent, Leuven, and Toronto, the BN model had an AUC of 0.77, 0.72, and 0.70, respectively. A SVM model based on the same variables had an overall worse performance (AUC 0.71, 0.68, and 0.69) especially in the Ghent set, which had the highest percentage of missing the important GTV size data. When only patients with complete data sets were considered, the BN and SVM model performed more alike. Conclusions: Within the limitations of this study, the hypothesis is supported that BN models are better at handling missing data than SVM models and are therefore more suitable for the medical domain. Future works have to focus on improving the BN performance by including more patients, more variables, and more diversity.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Coincident learning for unsupervised anomaly detection of scientific instruments

Abstract Anomaly detection is an important task for complex scientific experiments and other complex systems (e.g. industrial facilities, manufacturing), where failures in a sub-system can lead to lost data, poor performance, or even damage to components. While scientific facilities generate a wealth of data, labeled anomalies may be rare (or even nonexistent), and expensive to acquire. Unsupervised approaches are therefore common and typically search for anomalies either by distance or density of examples in the input feature space (or some associated low-dimensional representation). This paper presents a novel approach called coincident learning for anomaly detection (CoAD), which is specifically designed for multi-modal tasks and identifies anomalies based on coincident behavior across two different slices of the feature space. We define an unsupervised metric, F ^ β , out of analogy to the supervised classification F β statistic. CoAD uses F ^ β to train an anomaly detection algorithm on unlabeled data , based on the expectation that anomalous behavior in one feature slice is coincident with anomalous behavior in the other. The method is illustrated using a synthetic outlier data set and a MNIST-based image data set, and is compared to prior state-of-the-art on two real-world tasks: a metal milling data set and our motivating task of identifying RF station anomalies in a particle accelerator.

43 PARTICLE ACCELERATORS↗