Search NASA⌕ Search

SEARCH · Search NASA

Results for “outlier detection”

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 91 records · Page 5

Assessment of Outliers in Alloy Datasets Using Unsupervised Techniques

We report advancements in data analytics techniques have enabled complex, disparate datasets to be leveraged for alloy design. Identifying outliers in a dataset can reduce noise, identify erroneous and/or anomalous records, prevent overfitting, and improve model assessment and optimization. In this work, two alloy datasets (9-12% Cr ferritic martensitic steels, and austenitic stainless steels) have been assessed for outliers using unsupervised techniques and supplemented with domain knowledge. Principal component analysis and k-means clustering were applied to the data, and points were assessed as outliers based on their distance away from other points in the cluster and from other points in the dataset. The outlier characteristics were investigated to determine both cluster-specific and overall trends in the properties of the outlier points. The approach demonstrated here is extensible to other alloy datasets for outlier identification and evaluation to improve the reliability of machine learning and modeling predictions for advanced alloy design.

36 MATERIALS SCIENCE↗

On the accuracy of Hipparcos using binary stars as a calibration tool

Stellar binary systems, specifically those that present the most accurate available orbital elements, are a reliable tool to test the accuracy of astrometric observations. We selected all 35 binaries with these characteristics. Our objective is to provide standard uncertainties for the positions and parallaxes measured by Hipparcos relative to this trustworthy set, as well as to check supposed correlations between several parameters (measurement residuals, positions, magnitudes, and parallaxes). In addition, using the high-confidence subset of visual–spectroscopic binaries, we implemented a validation test of the Hipparcos trigonometric parallaxes of binary systems that allowed the evaluation of their reliability. Standard and non-standard statistical analysis techniques were applied in order to achieve well-founded conclusions. In particular, errors-in-variables models such as the total least-squares method were used to validate Hipparcos parallaxes by comparison with those obtained directly from the orbital elements. Previously, we executed Thompson's τ technique in order to detect suspected outliers in the data. Furthermore, several statistical hypothesis tests were carried out to verify if our results were statistically significant. A statistically significant trend indicating larger Hipparcos angular separations with respect to the reference values in 5.2 ± 1.4 mas was found at the 10{sup −8} significance level. Uncertainties in the polar coordinates θ and ρ of 1.°8 and 6.3 mas, respectively, were estimated for the Hipparcos observations of binary systems. We also verified that the parallaxes of binary systems measured in this mission are absolutely compatible with the set of orbital parallaxes obtained from the most accurate orbits at least at the 95% confidence level. This methodology allows us to better estimate the accuracy of Hipparcos observations of binary systems. Indeed, further application to the data collected by Gaia should yield a standard procedure to compare both data sets.

79 ASTRONOMY AND ASTROPHYSICS↗

A Multi-layer, Hierarchical Information Management System for the Smart Grid

This paper presents the modeling approach, methodologies, and initial results of setting up a multi-layer, hierarchical information management system (IMS) for the smart grid. The IMS allows its users to analyze the data collected by multiple control and communication networks to characterize the states of the smart grid. Abnormal, corrupted, or erroneous measurement data and outliers are detected and analyzed to identify whether they are caused by random equipment failures, unintentional human errors, or deliberate tempering attempts. Data collected from different information networks are crosschecked for data integrity based on redundancy, dependency, correlation, or cross-correlations, which reveal the interdependency between data sets. A hierarchically structured reasoning mechanism is used to rank possible causes of an event to aid the system operators to proactively respond or provide mitigation recommendations to remove or neutralize the threats. The model provides satisfactory performance on identifying the cause of an event and significantly reduces the need of processing myriads of data collected.

99 GENERAL AND MISCELLANEOUS↗

WE-D-BRA-01: FEATURED PRESENTATION and BEST IN PHYSICS (THERAPY): Predicting Potentially Problematic VMAT Treatment Plans Before Patient Specific QA Measurements

Purpose: Several promising IMRT QA tools have been developed in recent years to combat problems found in the lack of sensitivity in planar dose measurements analyzed using consensus gamma analysis criteria. The increased complexity and added information with such devices adds not only increased time, but new challenges in determining endpoints for pass/fail criteria. Using a large cohort of previously measured planar IMRT QA data, it may be possible to correlate potentially problematic plans with calculated plan metrics that can be done a priori, such that these tools can be used only in clinically relevant situations. Methods: 90 previously measured, clinically delivered VMAT plans were exported in DICOM RT format. Using a Matlab program, plan metrics were computed based on a previously developed set of equations (Du et al. 2014). These metrics included MU-weighted beam irregularity, which quantifies an MLC shape’s deviation from that of a circle. Machine delivery parameters such as MU delivered per degree and leaf movement in millimeters per MU were also calculated. Based on a previous analysis of 394 IMRT QA measurements, a “failing” plan was defined as one with less than 85% gamma pass rate when computed at 1% dose difference and 2 mm distance to agreement; 16 of the 90 plans were identified as failing Results: A ROC curve was generated with an AUC of 0.9409. All 16 outliers were detected with a specificity of 85.1% when using a threshold value based on a linear combination of the MU-weighted plan irregularity and the leaf speed in mm/degree (average MU delivered per degree multiplied by average leaf movement in mm per MU). Conclusion: This data supports that potentially problematic VMAT plans can possibly be predicted before measurement by assessing the MU-weighted irregularity of the MLC shapes combined with the averaged leaf speed of the MLC.

60 APPLIED LIFE SCIENCES↗

Metastatic Melanoma Induced Metabolic Changes in C57BL/6J Mouse Stomach Measured by 1H NMR Spectroscopy

Melanoma is a malignant tumor of melanocytes with high capability of invasion and rapid metastasis to other organs. Malignant melanoma is the most common metastatic malignancy found in gastrointestinal tract (GI). To the best of our knowledge, previous studies of melanoma in gastrointestinal tract are all clinical case reports. In this work, 1H NMR-based metabolomics approach is used to investigate the metabolite profiles differences of stomach tissue extracts of metastatic B16-F10 melanoma in C57BL/6J mouse and search for specific metabolite biomarker candidates. Principal Component Analysis (PCA), an unsupervised multivariate data analysis method, is used to detect possible outliers, while Orthogonal Projection to Latent Structure (OPLS), a supervised multivariate data analysis method, is employed to evaluate important metabolites responsible for discriminating the control and the melanoma groups. Both PCA and OPLS results reveal that the melanoma group can be well separated from its control group. Among the 50 identified metabolites, it is found that the concentrations of 19 metabolites are statistically and significantly changed with the levels of O-phosphocholine and hypoxanthine down-regulated while the levels of isoleucine, leucine, valine, isobutyrate, threonine, cadaverine, alanine, glutamate, glutamine, methionine, citrate, asparagine, tryptophan, glycine, serine, uracil, and formate up-regulated in the melanoma group. These significantly changed metabolites are associated with multiple biological pathways and may be potential biomarkers for metastatic melanoma in stomach.

59 BASIC BIOLOGICAL SCIENCES↗

Recent developments with the ORSER system

Additions to the ORSER remote sensing data processing package are described. The ORSER package consists of about 35 individual programs that are grouped into preprocessing, data analysis, and display subsystems. Additional data formats and data management, data transformation, and geometric correlation programs were supplemented to the preprocessing subsystem. Enhancements to the data analysis techniques include a maximum likelihood classifier (MAXCLASS) and a new version of the STATS program which makes delineation of training areas easier and allows for detection of outlier points. Ongoing developments are also described.

Baumer, G. M.↗

An Automated Algorithm to Screen Massive Training Samples for a Global Impervious Surface Classification

An algorithm is developed to automatically screen the outliers from massive training samples for Global Land Survey - Imperviousness Mapping Project (GLS-IMP). GLS-IMP is to produce a global 30 m spatial resolution impervious cover data set for years 2000 and 2010 based on the Landsat Global Land Survey (GLS) data set. This unprecedented high resolution impervious cover data set is not only significant to the urbanization studies but also desired by the global carbon, hydrology, and energy balance researches. A supervised classification method, regression tree, is applied in this project. A set of accurate training samples is the key to the supervised classifications. Here we developed the global scale training samples from 1 m or so resolution fine resolution satellite data (Quickbird and Worldview2), and then aggregate the fine resolution impervious cover map to 30 m resolution. In order to improve the classification accuracy, the training samples should be screened before used to train the regression tree. It is impossible to manually screen 30 m resolution training samples collected globally. For example, in Europe only, there are 174 training sites. The size of the sites ranges from 4.5 km by 4.5 km to 8.1 km by 3.6 km. The amount training samples are over six millions. Therefore, we develop this automated statistic based algorithm to screen the training samples in two levels: site and scene level. At the site level, all the training samples are divided to 10 groups according to the percentage of the impervious surface within a sample pixel. The samples following in each 10% forms one group. For each group, both univariate and multivariate outliers are detected and removed. Then the screen process escalates to the scene level. A similar screen process but with a looser threshold is applied on the scene level considering the possible variance due to the site difference. We do not perform the screen process across the scenes because the scenes might vary due to the phenology, solar-view geometry, and atmospheric condition etc. factors but not actual landcover difference. Finally, we will compare the classification results from screened and unscreened training samples to assess the improvement achieved by cleaning up the training samples. Keywords:

Tan, Bin↗

Optical Navigation Attitude Estimation and Calibration Performance Improvement using Outlier Rejection

Spacecraft optical navigation (OpNav) systems process a sequence of images of celestial bodies against a starfield background to estimate the position and velocity of the vehicle. While attitude is sometimes available from an onboard star tracker, it is often desirable to recognize the background stars in the OpNav images to better align the image. While many image processing algorithms exist for finding stars, efficiency and reliability remain key issues in the presence of extended bodies(e.g. the Moon, Earth), especially when attempting to solve the full lost-in-space problem. Some star outliers(stars identified with high residuals)could appear in the camera field of view, however using them in the attitude estimation or camera calibration would lead to less accurate results. Therefore, we require new and robust approaches to remove these outliers before any further processing. The emphasis of the work is on developing a simple and robust iterative technique to detect and reject the outliers which could be found in any frame during the lost in space attitude determination or during the camera calibration. These outliers are determined based on the residuals of the centroids of the detected stars and the corresponding location using the star catalog. If the residuals exceed a predetermined threshold value, the object will be detected as an outlier and will be removed before another attitude determination and calibration iteration is performed. The performance for both attitude determination and on-orbit camera calibration are improved by an almost two-fold increase in accuracy when applying this outlier rejection technique.

OpNav↗

SU-E-T-51: Bayesian Network Models for Radiotherapy Error Detection

Purpose: To develop a probabilistic model of radiotherapy plans using Bayesian networks that will detect potential errors in radiation delivery. Methods: Semi-structured interviews with medical physicists and other domain experts were employed to generate a set of layered nodes and arcs forming a Bayesian Network (BN) which encapsulates relevant radiotherapy concepts and their associated interdependencies. Concepts in the final network were limited to those whose parameters are represented in the institutional database at a level significant enough to develop mathematical distributions. The concept-relation knowledge base was constructed using the Web Ontology Language (OWL) and translated into Hugin Expert Bayes Network files via the the RHugin package in the R statistical programming language. A subset of de-identified data derived from a Mosaiq relational database representing 1937 unique prescription cases was processed and pre-screened for errors and then used by the Hugin implementation of the Estimation-Maximization (EM) algorithm for machine learning all parameter distributions. Individual networks were generated for each of several commonly treated anatomic regions identified by ICD-9 neoplasm categories including lung, brain, lymphoma, and female breast. Results: The resulting Bayesian networks represent a large part of the probabilistic knowledge inherent in treatment planning. By populating the networks entirely with data captured from a clinical oncology information management system over the course of several years of normal practice, we were able to create accurate probability tables with no additional time spent by experts or clinicians. These probabilistic descriptions of the treatment planning allow one to check if a treatment plan is within the normal scope of practice, given some initial set of clinical evidence and thereby detect for potential outliers to be flagged for further investigation. Conclusion: The networks developed here support the use of probabilistic models into clinical chart checking for improved detection of potential errors in RT plans.

60 APPLIED LIFE SCIENCES↗

The Harmonized Landsat and Sentinel-2 Surface Reflectance Data Set

The Harmonized Landsat and Sentinel-2 (HLS) project is a NASA initiative aiming to produce a VirtualConstellation (VC) of surface reflectance (SR) data acquired by the Operational Land Imager (OLI) and MultiSpectral Instrument (MSI) aboard Landsat 8 and Sentinel-2 remote sensing satellites, respectively. The HLS products are based on a set of algorithms to obtain seamless products from both sensors (OLI and MSI): atmospheric correction, cloud and cloud-shadow masking, spatial co-registration and common gridding, bidirectional reflectance distribution function normalization and spectral bandpass adjustment. Three products are derivedfrom the HLS processing chain: (i) S10: full resolution MSI SR at 10 m, 20 m and 60 m spatial resolutions; (ii)S30: a 30 m MSI Nadir BRDF (Bidirectional Reflectance Distribution Function)-Adjusted Reflectance (NBAR);(iii) L30: a 30 m OLI NBAR. All three products are processed for every Level-1 input products from Landsat 8/OLI (L1T) and Sentinel-2/MSI (L1C). As of version 1.3, the HLS data set covers 10.35 million km2 and spans from first Landsat 8 data (2013); Sentinel-2 data spans from October 2015. The L30 and S30 show a good consistency with coarse spatial resolution products, in particular MODIS Collection 6 MCD09CMG products (overall deviations do not exceed 11%) that are used as a reference for quality assurance. The spatial co-registration of the HLS is improved compared to original Landsat 8 L1T and Sentinel 2A L1C products, for which misregistration issues between multi-temporal data are known. In particular, the resulting computed circular errors at 90% for the HLS product are 6.2 m and 18.8 m, for S10 and L30 products, respectively. The main known issue of the current data set remains the Sentinel-2 cloud mask with many cloud detection omissions. The cross-comparison with MODIS was used to flag products with most evident non-detected clouds. A time series outlier filtering approach is suggested to detect remaining clouds. Finally, several time series are presented to highlight the high potential of the HLS data set for crop monitoring.

Landsat Sentinel-2↗

Automatic Feature Tracking on Small Bodies for Autonomous Approach

Abstract—The autonomous approach of a spacecraft to an asteroid or comet (a small body) relies heavily on visual feature tracking to aid in estimating relative trajectories and the properties of the small body. Feature tracking for small bodies brings several challenges, including changing lighting, poor visual texture, and a concentration of features in a small part of an image. Six existing, open-source algorithms for feature tracking were tested on a simulated dataset and compared to the ground truth in the path of features. The main finding is that none of the algorithms provide all of the desired characteristics of long feature tracks with low errors and few outliers. Instead, there is a trade-off between long feature tracks and low error. The feature-matching algorithms SIFT, and BRISK provide good error characteristics, but short feature tracks, whereas the optical flow algorithm KLT provides long feature tracks, but with many features of large error. Given the challenges in feature tracking, it is recommended to focus development on each component of a feature tracking system: detection, description, and outlier rejection.

Morrell, Benjamin J↗

SU-E-J-85: Leave-One-Out Perturbation (LOOP) Fitting Algorithm for Absolute Dose Film Calibration

Purpose: To introduce an outliers-recognition fitting routine for film dosimetry. It cannot only be flexible with any linear and non-linear regression but also can provide information for the minimal number of sampling points, critical sampling distributions and evaluating analytical functions for absolute film-dose calibration. Methods: The technique, leave-one-out (LOO) cross validation, is often used for statistical analyses on model performance. We used LOO analyses with perturbed bootstrap fitting called leave-one-out perturbation (LOOP) for film-dose calibration . Given a threshold, the LOO process detects unfit points (“outliers”) compared to other cohorts, and a bootstrap fitting process follows to seek any possibilities of using perturbations for further improvement. After that outliers were reconfirmed by a traditional t-test statistics and eliminated, then another LOOP feedback resulted in the final. An over-sampled film-dose- calibration dataset was collected as a reference (dose range: 0-800cGy), and various simulated conditions for outliers and sampling distributions were derived from the reference. Comparisons over the various conditions were made, and the performance of fitting functions, polynomial and rational functions, were evaluated. Results: (1) LOOP can prove its sensitive outlier-recognition by its statistical correlation to an exceptional better goodness-of-fit as outliers being left-out. (2) With sufficient statistical information, the LOOP can correct outliers under some low-sampling conditions that other “robust fits”, e.g. Least Absolute Residuals, cannot. (3) Complete cross-validated analyses of LOOP indicate that the function of rational type demonstrates a much superior performance compared to the polynomial. Even with 5 data points including one outlier, using LOOP with rational function can restore more than a 95% value back to its reference values, while the polynomial fitting completely failed under the same conditions. Conclusion: LOOP can cooperate with any fitting routine functioning as a “robust fit”. In addition, it can be set as a benchmark for film-dose calibration fitting performance.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Patterns of genomic variation in Coho salmon following reintroduction to the interior Columbia River

Abstract Coho salmon were extirpated in the mid‐20th century from the interior reaches of the Columbia River but were reintroduced with relatively abundant source stocks from the lower Columbia River near the Pacific coast. Reintroduction of Coho salmon to the interior Columbia River (Wenatchee River) using lower river stocks placed selective pressures on the new colonizers due to substantial differences with their original habitat such as migration distance and navigation of six additional hydropower dams. We used restriction site‐associated DNA sequencing ( RAD ‐seq) to genotype 5,392 SNP s in reintroduced Coho salmon in the Wenatchee River over four generations to test for signals of temporal structure and adaptive variation. Temporal genetic structure among the three broodlines of reintroduced fish was evident among the initial return years (2000, 2001, and 2002) and their descendants, which indicated levels of reproductive isolation among broodlines. Signals of adaptive variation were detected from multiple outlier tests and identified candidate genes for further study. This study illustrated that genetic variation and structure of reintroduced populations are likely to reflect source stocks for multiple generations but may shift over time once established in nature.

Campbell, Nathan R.↗

A Probabilistic Autoencoder for Type Ia Supernova Spectral Time Series

We construct a physically parameterized probabilistic autoencoder (PAE) to learn the intrinsic diversity of Type Ia supernovae (SNe Ia) from a sparse set of spectral time series. The PAE is a two-stage generative model, composed of an autoencoder that is interpreted probabilistically after training using a normalizing flow. We demonstrate that the PAE learns a low-dimensional latent space that captures the nonlinear range of features that exists within the population and can accurately model the spectral evolution of SNe Ia across the full range of wavelength and observation times directly from the data. By introducing a correlation penalty term and multistage training setup alongside our physically parameterized network, we show that intrinsic and extrinsic modes of variability can be separated during training, removing the need for the additional models to perform magnitude standardization. We then use our PAE in a number of downstream tasks on SNe Ia for increasingly precise cosmological analyses, including the automatic detection of SN outliers, the generation of samples consistent with the data distribution, and solving the inverse problem in the presence of noisy and incomplete data to constrain cosmological distance measurements. We find that the optimal number of intrinsic model parameters appears to be three, in line with previous studies, and show that we can standardize our test sample of SNe Ia with an rms of 0.091 ± 0.010 mag, which corresponds to 0.074 ± 0.010 mag if peculiar velocity contributions are removed.

79 ASTRONOMY AND ASTROPHYSICS↗

Searching for Novel Chemistry in Exoplanetary Atmospheres Using Machine Learning for Anomaly Detection

Abstract The next generation of telescopes will yield a substantial increase in the availability of high-quality spectroscopic data for thousands of exoplanets. The sheer volume of data and number of planets to be analyzed greatly motivate the development of new, fast, and efficient methods for flagging interesting planets for reobservation and detailed analysis. We advocate the application of machine learning (ML) techniques for anomaly (novelty) detection to exoplanet transit spectra, with the goal of identifying planets with unusual chemical composition and even searching for unknown biosignatures. We successfully demonstrate the feasibility of two popular anomaly detection methods (local outlier factor and one-class support vector machine) on a large public database of synthetic spectra. We consider several test cases, each with different levels of instrumental noise. In each case, we use receiver operating characteristic curves to quantify and compare the performance of the two ML techniques.

Astronomy & Astrophysics↗