Search NASA⌕ Search

SEARCH · Search NASA

Results for “pattern 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 181 records · Page 10

Probabilistic variable-length segmentation of protein sequences for discriminative motif discovery (DiMotif) and sequence embedding (ProtVecX)

In this paper, we present peptide-pair encoding (PPE), a general-purpose probabilistic segmentation of protein sequences into commonly occurring variable-length sub-sequences. The idea of PPE segmentation is inspired by the byte-pair encoding (BPE) text compression algorithm, which has recently gained popularity in subword neural machine translation. We modify this algorithm by adding a sampling framework allowing for multiple ways of segmenting a sequence. PPE segmentation steps can be learned over a large set of protein sequences (Swiss-Prot) or even a domain-specific dataset and then applied to a set of unseen sequences. This representation can be widely used as the input to any downstream machine learning tasks in protein bioinformatics. In particular, here, we introduce this representation through protein motif discovery and protein sequence embedding. (i) DiMotif: we present DiMotif as an alignment-free discriminative motif discovery method and evaluate the method for finding protein motifs in three different settings: (1) comparison of DiMotif with two existing approaches on 20 distinct motif discovery problems which are experimentally verified, (2) classification-based approach for the motifs extracted for integrins, integrin-binding proteins, and biofilm formation, and (3) in sequence pattern searching for nuclear localization signal. The DiMotif, in general, obtained high recall scores, while having a comparable F1 score with other methods in the discovery of experimentally verified motifs. Having high recall suggests that the DiMotif can be used for short-list creation for further experimental investigations on motifs. In the classification-based evaluation, the extracted motifs could reliably detect the integrins, integrin-binding, and biofilm formation-related proteins on a reserved set of sequences with high F1 scores. (ii) ProtVecX: we extend k-mer based protein vector (ProtVec) embedding to variablelength protein embedding using PPE sub-sequences. We show that the new method of embedding can marginally outperform ProtVec in enzyme prediction as well as toxin prediction tasks. In addition, we conclude that the embeddings are beneficial in protein classification tasks when they are combined with raw amino acids k-mer features.

59 BASIC BIOLOGICAL SCIENCES↗

Probabilistic variable-length segmentation of protein sequences for discriminative motif discovery (DiMotif) and sequence embedding (ProtVecX)

ABSTRACT In this paper, we present peptide-pair encoding (PPE), a general-purpose probabilistic segmentation of protein sequences into commonly occurring variable-length sub-sequences. The idea of PPE segmentation is inspired by the byte-pair encoding (BPE) text compression algorithm, which has recently gained popularity in subword neural machine translation. We modify this algorithm by adding a sampling framework allowing for multiple ways of segmenting a sequence. PPE segmentation steps can be learned over a large set of protein sequences (Swiss-Prot) or even a domain-specific dataset and then applied to a set of unseen sequences. This representation can be widely used as the input to any downstream machine learning tasks in protein bioinformatics. In particular, here, we introduce this representation through protein motif discovery and protein sequence embedding. (i) DiMotif: we present DiMotif as an alignment-free discriminative motif discovery method and evaluate the method for finding protein motifs in three different settings: (1) comparison of DiMotif with two existing approaches on 20 distinct motif discovery problems which are experimentally verified, (2) classification-based approach for the motifs extracted for integrins, integrin-binding proteins, and biofilm formation, and (3) in sequence pattern searching for nuclear localization signal. The DiMotif, in general, obtained high recall scores, while having a comparable F1 score with other methods in the discovery of experimentally verified motifs. Having high recall suggests that the DiMotif can be used for short-list creation for further experimental investigations on motifs. In the classification-based evaluation, the extracted motifs could reliably detect the integrins, integrin-binding, and biofilm formation-related proteins on a reserved set of sequences with high F1 scores. (ii) ProtVecX: we extend k-mer based protein vector (ProtVec) embedding to variable-length protein embedding using PPE sub-sequences. We show that the new method of embedding can marginally outperform ProtVec in enzyme prediction as well as toxin prediction tasks. In addition, we conclude that the embeddings are beneficial in protein classification tasks when they are combined with raw k-mer features. Availability Implementations of our method will be available under the Apache 2 licence at http://llp.berkeley.edu/dimotif and http://llp.berkeley.edu/protvecx .

Asgari, Ehsaneddin↗

Convolutional Neural Networks for Hydrometeor Classification using Dual Polarization Doppler Radars

Traditional fuzzy logic hydrometeor classification algorithm is a common way to classify precipitation type from dual polarization doppler radar. We propose a deep learning-based method to estimate hydrometeors efficiently using observed radar variables such as horizontal reflectivity (Z H ), differential reflectivity (Z DR ), correlation coefficient (ρ HV ) and specific differential phase (K DP ) from National Weather Service NEXRAD collected at Vance AFB facility at the first elevation angle from January 1st, 2015 to July 31th, 2019. We stack matrices of values from these four polarimetric variables as one 3D array. Samples are preprocessed and divided into training, validation and test set with four target hydrometeor categories (Ice Crystals (IC), Dry Snow (DS), Light and/or Moderate Rain (RA) and Big Drops (rain) (BD)). We developed and optimized five Convolutional Neural Networks (CNNs) architectures and achieved an accuracy of 87.23% and 93.736% respectively using modified ResNet with two different input data selection strategies for hydrometeor classification. Training data selection strategies were important to ensure use of available samples in training for robust performance evaluated by applying the models to novel time period beyond what was use to train the model. Seasonal variation in atmospheric conditions lead to seasonal patterns of liquid vs solid forms of precipitation, that poses challenge for classifier and offer insights into domain specific approaches required for problem of hydrometeor identification. Computationally efficient and scalable approach for classification of hydrometeors offer opportunties to effectively use the large volumes of rich time series of radar observations that are becoming increasingly available.

Lu, Yuping↗

Using Deep Learning for Automated Communication Pattern Characterization: Little Steps and Big Challenges

Characterization of a parallel application’s communication patterns can be useful for performance analysis, debugging, and system design. However, obtaining and interpreting a characterization can be difficult. AChax implements an approach that uses search and a library of known communication patterns to automatically characterize communication patterns. Our approach has some limitations that reduce its effectiveness for the patterns and pattern combinations used by some real-world applications. By viewing AChax’s pattern recognition problem as an image recognition problem, it may be possible to use deep learning to address these limitations. In this position paper, we present our current ideas regarding the benefits and challenges of integrating deep learning into AChax and our conclusion that a hybrid approach combining deep learning classification, regression, and the existing AChax approach may be the best long-term solution to the problem of parameterizing recognized communication patterns.

Roth, Philip↗

Atomic resolution convergent beam electron diffraction analysis using convolutional neural networks

Two types of convolutional neural network (CNN) models, a discrete classification network and a continuous regression network, were trained to determine local sample thickness from convergent beam diffraction (CBED) patterns of SrTiO 3 collected in a scanning transmission electron microscope (STEM) at atomic column resolution. Acquisition of atomic resolution CBED patterns for this purpose requires careful balancing of CBED feature size in pixels, acquisition speed, and detector dynamic range. The training datasets were derived from multislice simulations, which must be convolved with incoherent source broadening. Sample thicknesses were also determined using quantitative high-angle annular dark-field (HAADF) STEM images acquired simultaneously. The regression CNN performed well on sample thinner than 35 nm, with 70% of the CNN results within 1 nm of HAADF thickness, and 1.0 nm overall root mean square error between the two measurements. The classification CNN was trained for a thicknesses up to 100 nm and yielded 66% of CNN results within one classification increment of 2 nm of HAADF thickness. Our approach depends on methods from computer vision including transfer learning and image augmentation.

36 MATERIALS SCIENCE↗

Classification of genes based on gene expression analysis

Systems biology and bioinformatics are now major fields for productive research. DNA microarrays and other array technologies and genome sequencing have advanced to the point that it is now possible to monitor gene expression on a genomic scale. Gene expression analysis is discussed and some important clustering techniques are considered. The patterns identified in the data suggest similarities in the gene behavior, which provides useful information for the gene functionalities. We discuss measures for investigating the homogeneity of gene expression data in order to optimize the clustering process. We contribute to the knowledge of functional roles and regulation of E. coli genes by proposing a classification of these genes based on consistently correlated genes in expression data and similarities of gene expression patterns. A new visualization tool for targeted projection pursuit and dimensionality reduction of gene expression data is demonstrated.

60 APPLIED LIFE SCIENCES↗

Integrated Genomic Analysis Identifies Clinically Relevant Subtypes of Glioblastoma Characterized by Abnormalities in PDGFRA, IDH1, EGFR, and NF1

The Cancer Genome Atlas Network recently cataloged recurrent genomic abnormalities in glioblastoma multiforme (GBM). We describe a robust gene expression-based molecular classification of GBM into Proneural, Neural, Classical, and Mesenchymal subtypes and integrate multidimensional genomic data to establish patterns of somatic mutations and DNA copy number. Aberrations and gene expression of EGFR, NF1, and PDGFRA/IDH1 each define the Classical, Mesenchymal, and Proneural subtypes, respectively. Gene signatures of normal brain cell types show a strong relationship between subtypes and different neural lineages. Additionally, response to aggressive therapy differs by subtype, with the greatest benefit in the Classical subtype and no benefit in the Proneural subtype. We provide a framework that unifies transcriptomic and genomic dimensions for GBM molecular stratification with important implications for future studies.

59 BASIC BIOLOGICAL SCIENCES↗

Machine learning for interpreting coherent X-ray speckle patterns

Speckle patterns produced by coherent X-ray have a close relationship with the internal structure of materials but quantitative inversion of the relationship to determine structure from speckle patterns is challenging. Here, in this work, we investigate the link between coherent X-ray speckle patterns and sample structures using a model 2D disk system and explore the ability of machine learning to learn aspects of the relationship. Specifically, we train a deep neural network to classify the coherent X-ray speckle patterns according to the disk number density in the corresponding structure. It is demonstrated that the classification system is accurate for both non-disperse and disperse size distributions.

36 MATERIALS SCIENCE↗

Effect of Diel Cycling Temperature, Relative Humidity, and Synthetic Route on the Surface Morphology and Hydrolysis of α -U 3 O 8

The speciation and morphological changes of α-U 3 O 8 following aging under diel cycling temperature and relative humidity (RH) have been examined. This work advances the knowledge of U-oxide hydration as a result of synthetic route and environmental conditions, ultimately giving novel insight into nuclear material provenance. α-U 3 O 8 was synthesized via the washed uranyl peroxide (UO 4 ) and ammonium uranyl carbonate (AUC) synthetic routes to produce unaged starting materials with different morphologies. α-U 3 O 8 from UO 4 is comprised of subrounded particles, while α-U 3 O 8 from AUC contains blocky, porous particles approximately an order of magnitude larger than particles from UO 4 . For aging, a humidity chamber was programmed for continuous daily cycles of 12 “high” hours of 45 °C and 90% RH, and 12 “low” hours of 25 °C and 20% RH. Samples were analyzed at varying intervals of 14, 24, 36, 43, and 54 days. At each aging interval, crystallographic changes were measured via powder X-ray diffraction coupled with whole pattern fitting for quantitative analysis. Morphologic effects were studied via scanning electron microscopy and 12-way classification via machine learning. While all samples were found to have distinguishing morphologic characteristics (93.2% classification accuracy), α-U 3 O 8 from UO 4 had more apparent change with increasing aging time. Nonetheless, α-U 3 O 8 from AUC was found to hydrate more quickly than α-U 3 O 8 from UO 4 , which can likely be attributed to its larger surface area and porous starting material morphology.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

VLSI Cells Placement Using the Neural Networks

The artificial neural networks have been studied for several years. Their effectiveness makes it possible to expect high performances. The privileged fields of these techniques remain the recognition and classification. Various applications of optimization are also studied under the angle of the artificial neural networks. They make it possible to apply distributed heuristic algorithms. In this article, a solution to placement problem of the various cells at the time of the realization of an integrated circuit is proposed by using the KOHONEN network.

99 GENERAL AND MISCELLANEOUS↗

Neuroevolution of Spiking Neural Networks Using Compositional Pattern Producing Networks

Spiking neural networks (SNNs) offer tremendous potential for the future of AI, including the ability to be implemented efficiently on neuromorphic systems. One of the challenges in building functioning SNNs is the training process, as standard error back-propagation cannot be easily applied. In this work, we extend an evolutionary approach for training SNNs by implementing an indirect encoding of individuals. Specifically, we evolve SNNs using Compositional Pattern Producing Networks, which are able to learn the connectivity patterns between neurons defined in a coordinate space. We validate the approach on multiple control and classification tasks.

Elbrecht, Daniel↗

Automatic Microaneurysm Detection and Characterization Through Digital Color Fundus Images

Ocular fundus images can provide information about retinal, ophthalmic, and even systemic diseases such as diabetes. Microaneurysms (MAs) are the earliest sign of Diabetic Retinopathy, a frequently observed complication in both type 1 and type 2 diabetes. Robust detection of MAs in digital color fundus images is critical in the development of automated screening systems for this kind of disease. Automatic grading of these images is being considered by health boards so that the human grading task is reduced. In this paper we describe segmentation and the feature extraction methods for candidate MAs detection.We show that the candidate MAs detected with the methodology have been successfully classified by a MLP neural network (correct classification of 84percent).

97 MATHEMATICS AND COMPUTING↗

Pelvic Arterial Anatomy Relevant to Prostatic Artery Embolisation and Proposal for Angiographic Classification

PurposeTo describe and categorize the angiographic findings regarding prostatic vascularization, propose an anatomic classification, and discuss its implications for the PAE procedure.MethodsAngiographic findings from 143 PAE procedures were reviewed retrospectively, and the origin of the inferior vesical artery (IVA) was classified into five subtypes as follows: type I: IVA originating from the anterior division of the internal iliac artery (IIA), from a common trunk with the superior vesical artery (SVA); type II: IVA originating from the anterior division of the IIA, inferior to the SVA origin; type III: IVA originating from the obturator artery; type IV: IVA originating from the internal pudendal artery; and type V: less common origins of the IVA. Incidences were calculated by percentage.ResultsTwo hundred eighty-six pelvic sides (n = 286) were analyzed, and 267 (93.3 %) were classified into I–IV types. Among them, the most common origin was type IV (n = 89, 31.1 %), followed by type I (n = 82, 28.7 %), type III (n = 54, 18.9 %), and type II (n = 42, 14.7 %). Type V anatomy was seen in 16 cases (5.6 %). Double vascularization, defined as two independent prostatic branches in one pelvic side, was seen in 23 cases (8.0 %).ConclusionsDespite the large number of possible anatomical variations of male pelvis, four main patterns corresponded to almost 95 % of the cases. Evaluation of anatomy in a systematic fashion, following a standard classification, will make PAE a faster, safer, and more effective procedure.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Histologic Subtype in Core Lung Biopsies of Early-Stage Lung Adenocarcinoma is a Prognostic Factor for Treatment Response and Failure Patterns After Stereotactic Body Radiation Therapy

Purpose: Stereotactic body radiation therapy (SBRT) has emerged as an effective treatment for early-stage lung cancer. The histologic subtype of surgically resected lung adenocarcinoma is recognized as a prognostic factor, with the presence of solid or micropapillary patterns predicting poor outcomes. We describe the outcomes after SBRT for early-stage lung adenocarcinoma stratified by histologic subtype. Methods and Materials: We identified 119 consecutive patients (124 lesions) with stage I to IIA lung adenocarcinoma who had undergone definitive SBRT at our institution from August 2008 to August 2015 and had undergone core biopsy. Histologic subtyping was performed according to the 2015 World Health Organization classification. Of the 124 tumors, 37 (30%) were a high-risk subtype, defined as containing a component of solid and/or micropapillary pattern. The cumulative incidences of local, nodal, regional, and distant failure were compared between the high-risk and non–high-risk adenocarcinoma subtypes using Gray's test, and multivariable-adjusted hazard ratios (HRs) were estimated from propensity score–weighted Cox regression models. Results: The median follow-up for the entire cohort was 17 months and for surviving patients was 21 months. The 1-year cumulative incidence of and adjusted HR for local, nodal, regional, and distant failure in high-risk versus non–high-risk lesions was 7.3% versus 2.7% (HR 16.8; 95% confidence interval [CI] 3.5-81.4), 14.8% versus 2.6% (HR 3.8; 95% CI 0.95-15.0), 4.0% versus 1.2% (HR 20.9; 95% CI 2.3-192.3), and 22.7% versus 3.6% (HR 6.9; 95% CI 2.2-21.1), respectively. No significant difference was seen with regard to overall survival. Conclusions: The outcomes after SBRT for early-stage adenocarcinoma of the lung correlate highly with histologic subtype, with micropapillary and solid tumors portending significantly higher rates of locoregional and metastatic progression. In this context, the histologic subtype determined from core biopsies is a prognostic factor and could have important implications for patient selection, adjuvant treatment, biopsy methods, and clinical trial design.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

A probabilistic model for the identification of confinement regimes and edge localized mode behavior, with implications to scaling laws

Pattern recognition is becoming an important tool in fusion data analysis. However, fusion diagnostic measurements are often affected by considerable statistical uncertainties, rendering the extraction of useful patterns a significant challenge. Therefore, we assume a probabilistic model for the data and perform pattern recognition in the space of probability distributions. We show the considerable advantage of our method for identifying confinement regimes and edge localized mode behavior, and we discuss the potential for scaling laws.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Materials data analytics for 9% Cr family steel

A materials data analytics (MDA) methodology was developed in this study to evaluate publicly available information on 9% Cr family steel and to handle nonlinear relationships and the sparsity in materials data for this alloy class. The overarching goal is to accelerate the design process as well as to reduce the time and expense associated with qualification testing of new alloys for fossil energy applications. Data entries in the analyzed data set for 82 iron–base alloy compositions, several processing parameters, and results of tensile mechanical tests selected for this study were arranged in 34 columns by 915 rows. While detailed microstructural information was not available, it is assumed that the compositional space for the 9 to 12% Cr steels is limited such that all data entries have a tempered martensitic microstructure during service. Establishing a hierarchy of first–order trends in the publicly available data requires the MDA to filter out the biases. Complexity of the phase transformations and microstructure evolution in the multicomponent alloys (using 21 chemical elements) with major influence on mechanical properties, leads to inefficiency in direct application of unbiased linear regression across the entire data space. To address the nonlinearity, analyses of tensile data were performed in composition–based clusters. Clusters corresponding to moderately frequent patterns and maximized information gain were further refined by using p–norm distance measures, matching the alloy classification groups adopted by industry. Lastly, the evolutionary method of propagating an ensemble of competing cluster–based models proved to be a viable option in dealing with scarce, multidimensional data.

36 MATERIALS SCIENCE↗

Organochlorine pesticides (OCPs) and polychlorinated biphenyls (PCBs) in Cyprinidae fish: Towards hints of their arrangements using advanced classification methods

Highlights: • SOM and DT classifiers searched for relation among 25 POPs in three Cyprinidae fish. • The concentrations of indicator congeners dominated over dioxin-like PCBs. • Based on SOM, both dioxin-like and indicator PCBs differentiated the fish species. • DT assorted the samples on the levels of PCB-126, -156, -170, -28, -105, and -114. To tackle the ever-present global concern regarding human exposure to persistent organic pollutants (POPs) via food products, this study strived to indicate associations between organochlorine pesticides (OCPs) and polychlorinated biphenyls (PCBs) in lake-fish tissue depending on the species and sampling season. Apart from the monitoring initiatives recommended in the Global Monitoring Plan for POPs, the study discussed 7 OCPs and 18 PCB congeners determined in three Cyprinidae species (rudd, carp, and Prussian carp) from Vransko Lake (Croatia), which are widely domesticated and reared as food fish across Europe and Asia. We exploit advanced classification algorithms, the Kohonen self-organizing maps (SOM) and Decision Trees (DT), to search for POP patterns typical for the investigated species. As indicated by SOM, some of the dioxin-like and non-dioxin-like PCBs (PCB-28, PCB-74, PCB-52, PCB-101, PCB-105, PCB-114, PCB-118, PCB-156 and PCB-157), α-HCH and β-HCH caused dissimilarities among fish species, but regardless of their weight and length. To support these suggestions, DT analysis sequenced the fish species and seasons based on the concentration of heavier congeners. The presented assumptions indicated that the supplemental application of SOM and DT offers advantageous features over the usually rough interpretation of POPs pattern and over the single use of the methods.

54 ENVIRONMENTAL SCIENCES↗

EXTENDED NARROW-LINE EMISSION IN THE BRIGHT SEYFERT 1.5 GALAXY HE 2211-3903

Extended narrow-line regions (ENLRs) and extended emission-line regions have been the focus of integral field spectroscopy aiming at the inner kiloparsecs of nearby Seyfert galaxies as well as the larger environment of high-redshift QSOs. Based on observations with the Wide Field Spectrograph at the 2.3 m telescope of the Australian National University, we present spatially resolved emission-line diagnostics of the bright Seyfert 1.5 galaxy HE 2211-3903 which is drawn from a sample of the brightest Seyfert galaxies at z < 0.06 with luminosities around the classical Seyfert/QSO demarcation. In addition to the previously known spiral arms of HE 2211-3903, the emission-line maps reveal a large-scale ring with a radius of about 6 kpc which is connected to the active galactic nucleus (AGN) through a bar-like structure. The overall gas kinematics indicates a disk rotation pattern. The emission-line ratios show Seyfert-type, H II region-type, and composite classifications, while there is no strong evidence of LINER-type ratios. Shock ionization is likely to be negligible throughout the galaxy. The composite line ratios are explained via a mixing line between AGN and H II region photoionization. Composite line ratios are predominantly found in between the H II regions in the circum-nuclear region, the bar-like structure to the east of the nucleus, and the eastern half of the ring, suggesting AGN photoionization of the low-density interstellar medium in an ENLR on galaxy scales. The line ratios in the nucleus indicate N enrichment, which is discussed in terms of chemical enrichment by Wolf-Rayet and asymptotic giant branch stars during past and ongoing nuclear starburst activity.

79 ASTRONOMY AND ASTROPHYSICS↗