Search NASA⌕ Search

SEARCH · Search NASA

Results for “clustering methods”

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 217 records · Page 12

A two-stage method for microcalcification cluster segmentation in mammography by deformable models

Purpose: Segmentation of microcalcification (MC) clusters in x-ray mammography is a difficult task for radiologists. Accurate segmentation is prerequisite for quantitative image analysis of MC clusters and subsequent feature extraction and classification in computer-aided diagnosis schemes. Methods: In this study, a two-stage semiautomated segmentation method of MC clusters is investigated. The first stage is targeted to accurate and time efficient segmentation of the majority of the particles of a MC cluster, by means of a level set method. The second stage is targeted to shape refinement of selected individual MCs, by means of an active contour model. Both methods are applied in the framework of a rich scale-space representation, provided by the wavelet transform at integer scales. Segmentation reliability of the proposed method in terms of inter and intraobserver agreements was evaluated in a case sample of 80 MC clusters originating from the digital database for screening mammography, corresponding to 4 morphology types (punctate: 22, fine linear branching: 16, pleomorphic: 18, and amorphous: 24) of MC clusters, assessing radiologists’ segmentations quantitatively by two distance metrics (Hausdorff distance—HDIST{sub cluster}, average of minimum distance—AMINDIST{sub cluster}) and the area overlap measure (AOM{sub cluster}). The effect of the proposed segmentation method on MC cluster characterization accuracy was evaluated in a case sample of 162 pleomorphic MC clusters (72 malignant and 90 benign). Ten MC cluster features, targeted to capture morphologic properties of individual MCs in a cluster (area, major length, perimeter, compactness, and spread), were extracted and a correlation-based feature selection method yielded a feature subset to feed in a support vector machine classifier. Classification performance of the MC cluster features was estimated by means of the area under receiver operating characteristic curve (Az ± Standard Error) utilizing tenfold cross-validation methodology. A previously developed B-spline active rays segmentation method was also considered for comparison purposes. Results: Interobserver and intraobserver segmentation agreements (median and [25%, 75%] quartile range) were substantial with respect to the distance metrics HDIST{sub cluster} (2.3 [1.8, 2.9] and 2.5 [2.1, 3.2] pixels) and AMINDIST{sub cluster} (0.8 [0.6, 1.0] and 1.0 [0.8, 1.2] pixels), while moderate with respect to AOM{sub cluster} (0.64 [0.55, 0.71] and 0.59 [0.52, 0.66]). The proposed segmentation method outperformed (0.80 ± 0.04) statistically significantly (Mann-Whitney U-test, p < 0.05) the B-spline active rays segmentation method (0.69 ± 0.04), suggesting the significance of the proposed semiautomated method. Conclusions: Results indicate a reliable semiautomated segmentation method for MC clusters offered by deformable models, which could be utilized in MC cluster quantitative image analysis.

60 APPLIED LIFE SCIENCES↗

High-performance dynamic quantum clustering on graphics processors

Clustering methods in machine learning may benefit from borrowing metaphors from physics. Dynamic quantum clustering associates a Gaussian wave packet with the multidimensional data points and regards them as eigenfunctions of the Schroedinger equation. The clustering structure emerges by letting the system evolve and the visual nature of the algorithm has been shown to be useful in a range of applications. Furthermore, the method only uses matrix operations, which readily lend themselves to parallelization. In this paper, we develop an implementation on graphics hardware and investigate how this approach can accelerate the computations. We achieve a speedup of up to two magnitudes over a multicore CPU implementation, which proves that quantum-like methods and acceleration by graphics processing units have a great relevance to machine learning.

97 MATHEMATICS AND COMPUTING↗

Thermodynamics of Metal Nanoparticles: Energies and Enthalpies of Formation of Magnesium Clusters and Nanoparticles as Large as 1.3 nm

The major obstacle that prevents reliable electronic structure studies of nanoparticles is the rapid increasing computational cost for benchmark calculations using coupled-cluster methods. We show that a CCSD(T) scheme with an MP2/CBS correction can reproduce accurate cohesive energies for magnesium clusters, and this scheme is much less computationally demanding than other reliable methods, so it is applied to Mg n with n up to 19, which enters the realm of nanoparticles. (The diameters of all Mg clusters n ≥ 11 are >1 nm). With the extended benchmark data, we validate exchange–correlation functionals into the nanoparticle regime and use the two best-validated functionals to calculate the enthalpy of formation of Mg 28 , with a diameter of 1.30 nm. We also calculated the enthalpy of formation of all Mg clusters and nanoparticles from Mg 2 to Mg 19 . This kind of reliable thermodynamic data on size-selected metal nanoparticles has been hard to come by, either experimentally or theoretically, but it is badly needed to support applications in catalysis, electrochemistry, and other technologies.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Agent-based method for distributed clustering of textual information

A computer method and system for storing, retrieving and displaying information has a multiplexing agent (20) that calculates a new document vector (25) for a new document (21) to be added to the system and transmits the new document vector (25) to master cluster agents (22) and cluster agents (23) for evaluation. These agents (22, 23) perform the evaluation and return values upstream to the multiplexing agent (20) based on the similarity of the document to documents stored under their control. The multiplexing agent (20) then sends the document (21) and the document vector (25) to the master cluster agent (22), which then forwards it to a cluster agent (23) or creates a new cluster agent (23) to manage the document (21). The system also searches for stored documents according to a search query having at least one term and identifying the documents found in the search, and displays the documents in a clustering display (80) of similarity so as to indicate similarity of the documents to each other.

Potok, Thomas E↗

The General Atomic and Molecular Electronic Structure System (GAMESS): Novel Methods on Novel Architectures

The primary focus of GAMESS over the last 5 years has been the development of new high-performance codes that are able to take effective and efficient advantage of the most advanced computer architectures, both CPU and accelerators. These efforts include employing density fitting and fragmentation methods to reduce the high scaling of well-correlated (e.g., coupled-cluster) methods as well as developing novel codes that can take optimal advantage of graphical processing units and other modern accelerators. Because accurate wave functions can be very complex, an important new functionality in GAMESS is the quasi-atomic orbital analysis, an unbiased approach to the understanding of covalent bonds embedded in the wave function. Finally, best practices for the maintenance and distribution of GAMESS are also discussed.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Basis set convergence of the binding energies of strongly hydrogen-bonded atmospheric clusters

We investigate the basis set convergence of second order Møller Plesset perturbation theory for the binding energies of 11 strongly hydrogen-bonded clusters relevant to the atmosphere. The binding energies are calculated both as the uncorrected (UC) value, as well as by employing the counterpoise (CP) and the Same Number Of Optimized Parameters (SNOOP) correction schemes, with and without explicit correlation (F12). We find that the use of the F12 corrections is of utter importance for obtaining converged binding energies. Without F12 corrections at least a quadruple-ζ basis set is generally required to yield errors below 1 kcal mol -1 compared to the complete basis set (CBS) limit. Using coupled cluster methods we obtain a best estimate of the CCSD(T) CBS limit of the binding energies of the considered clusters and compare it with approximate DLPNO-CCSD(T) and DFT methods. We identify a pragmatic approach, relying solely on a series of double-ζ basis set calculations, for obtaining results in remarkably good agreement with our CBS estimate.

Chemistry↗

Partitioning of Large-Scale Power Electronics-Based Power Systems for Small-Signal Stability Analysis

The nodal admittance matrix (NAM)-based approach is suitable for analyzing the small-signal stability of large-scale power electronics-based power systems (PEPSs) as it preserves the system structure by utilizing the admittance matrix. Previously, NAM-based area partition has been proposed, which divides the system into various subareas and interconnections for easier analysis of the low-dimension matrix compared to the entire system-based high-dimension matrix. However, no partition algorithm has been presented for the NAM-based area partition method. This paper focuses on implementing the spectral partitioning algorithm for partitioning large-scale PEPSs into a low-dimension matrix to reduce the computation complexity of the analysis. These spectral components facilitate data transformation into a new space, enabling the application of traditional clustering methods like k-means. To evaluate the performance of the partitioning method, the subareas and interconnections obtained from the spectral clustering algorithm are incorporated into the NAM-based area partition method for a large system with 140 buses. The computational times of the original method, where the NAM-based criterion is directly applied to the entire system, are compared with those of the NAM-based partition method in MATLAB. PSCAD simulations of the whole system and the obtained subareas are conducted to validate the effectiveness of the proposed algorithm.

Nupur, Nupur↗

DCA++

With the DCA++ project, we aim to gain insight into the fascinating physics of strongly correlated electron systems by employing modern quantum cluster methods. The DCA++ code provides a state of the art implementation of the dynamical cluster approximation (DCA) and its DCA+ extension. High scalability and portable performance allow to exploit today's leadership computing systems.

SciDAC↗

Computational studies of metal-metal and metal-ligand interactions

Accurate calculations on the bond length, dipole moment, and harmonic frequency of CO are presented, using large basis sets and high levels of electron correlation. The geometric structure, force constants and binding energies of Cr(C0)6 and Cr(CO)5 are computed using large basis sets and high levels of electron correlation. The molecule 04(+) is studied, using large basis sets and high levels of electron correlation, including the CASSCF, CASSI and CASPT2 methods. Binding energies, geometries and frequencies are computed. Symmetry breaking is a particular problem for the antisymmetric stretch, which is addressed using the CASSI method. The symmetry breaking problem in 04(+) has also been studied using the Brueckner coupled cluster method. This gives results in good agreement with CASSI. A multi-region numerical integration scheme is investigated for use in Density Functional Calculations. This scheme is found to give comparable results to a widely used scheme based on the Euler-Maclaurin technique.

Barnes, Leslie A.↗

Development and optimization of SPECT gated blood pool cluster analysis for the prediction of CRT outcome

Purpose: Phase analysis of single photon emission computed tomography (SPECT) radionuclide angiography (RNA) has been investigated for its potential to predict the outcome of cardiac resynchronization therapy (CRT). However, phase analysis may be limited in its potential at predicting CRT outcome as valuable information may be lost by assuming that time-activity curves (TAC) follow a simple sinusoidal shape. A new method, cluster analysis, is proposed which directly evaluates the TACs and may lead to a better understanding of dyssynchrony patterns and CRT outcome. Cluster analysis algorithms were developed and optimized to maximize their ability to predict CRT response. Methods: About 49 patients (N = 27 ischemic etiology) received a SPECT RNA scan as well as positron emission tomography (PET) perfusion and viability scans prior to undergoing CRT. A semiautomated algorithm sampled the left ventricle wall to produce 568 TACs from SPECT RNA data. The TACs were then subjected to two different cluster analysis techniques, K-means, and normal average, where several input metrics were also varied to determine the optimal settings for the prediction of CRT outcome. Each TAC was assigned to a cluster group based on the comparison criteria and global and segmental cluster size and scores were used as measures of dyssynchrony and used to predict response to CRT. A repeated random twofold cross-validation technique was used to train and validate the cluster algorithm. Receiver operating characteristic (ROC) analysis was used to calculate the area under the curve (AUC) and compare results to those obtained for SPECT RNA phase analysis and PET scar size analysis methods. Results: Using the normal average cluster analysis approach, the septal wall produced statistically significant results for predicting CRT results in the ischemic population (ROC AUC = 0.73;p < 0.05 vs. equal chance ROC AUC = 0.50) with an optimal operating point of 71% sensitivity and 60% specificity. Cluster analysis results were similar to SPECT RNA phase analysis (ROC AUC = 0.78, p = 0.73 vs cluster AUC; sensitivity/specificity = 59%/89%) and PET scar size analysis (ROC AUC = 0.73, p = 1.0 vs cluster AUC; sensitivity/specificity = 76%/67%). Conclusions: A SPECT RNA cluster analysis algorithm was developed for the prediction of CRT outcome. Cluster analysis results produced results equivalent to those obtained from Fourier and scar analysis.

60 APPLIED LIFE SCIENCES↗

Quantum Chemistry via Walks in Determinant Space

There are many chemical questions of practical interest to the DOE that could be answered if there were an electronic structure method that provided consistently accurate results for all systems at an affordable computational cost. The coupled cluster method with single, double and perturbative triple excitations (CCSD(T)) is the most frequently used high-order method, but it has known deficiencies, e.g., in the description of stretched bonds. The full configuration interaction (FCI) method is the most robust method for treating electronic correlations, but it is little used because its computational cost scales exponentially in the size of the system. The largest calculation that has been done to date employed 10 billion determinants. In this regard, there was a major advance in 2010. The Alavi group at Cambridge University developed a stochastic approach to FCI --- combining it with ideas from quantum Monte Carlo (QMC) --- called FCIQMC, that allows one to go to a far larger number of determinants in certain circumstances. The computational cost is exponential in the system and basis size but with a much reduced exponent compared to conventional FCI. In this project Umrigar's group made several major improvements to the FCIQMC method that increased its efficiency by many orders of magnitude. In addition this project resulted in a cross-fertilization of ideas between the FCIQMC method, the older phaseless auxilliary-field quantum Monte Carlo (AFQMC) method developed by Zhang and Krakauer (two of the PI's of this project), and symmetry-restored wavefunctions developed by Scuseria (also a PI of this project).

36 MATERIALS SCIENCE↗

Computational studies of metal-metal and metal-ligand interactions

Accurate calculations on the bond length, dipole moment, and harmonic frequency of CO are presented, using large basis sets and high levels of electron correlation. The geometric structure, force constants and binding energies of Cr(CO)6 and Cr(CO)5 are computed using large basis sets and high levels of electron correlation. The molecule O4(+) is studied, using large basis sets and high levels of electron correlation, including the CASSCF, CASSI,and CASPT2 methods. Binding energies, geometries and frequencies are computed. Symmetry breaking is a particular problem for the antisymmetric stretch, which is addressed using the CASSI method. The symmetry breaking problem in O4(+) has also been studied using the Brueckner coupled-cluster method. This gives results in good agreement with CASSI. A multi-region numerical integration scheme is investigated for use in Density Functional Calculations. This scheme is found to give comparable results to a widely used scheme based on the Euler-Maclaurin technique.

Barnes, Leslie A.↗

Removal of spurious data in Bragg coherent diffraction imaging: an algorithm for automated data preprocessing

Bragg coherent diffraction imaging (BCDI) provides a powerful tool for obtaining high-resolution structural information from nanocrystalline materials. Here a BCDI sample consisting of a large number of randomly oriented nanoscale crystals is considered. Ideally, only one crystal is oriented to produce a Bragg peak on the detector. However, diffraction from other crystals often produces additional signals on the detector. Before the measured diffraction patterns can be processed into structural images, scientists routinely need to manually identify and remove the `alien' intensities from sources other than the intended crystal. With the development of modern high-coherence storage rings, such as the upgraded Advanced Photon Source (APS), the already slow process of manual preprocessing will be untenable for the large volumes of data that will be produced. An automated method of identifying and deleting alien intensities is proposed. This method exploits the fact that BCDI of a perfect crystal produces diffraction data with inversion symmetry around the Bragg peak. This approach uses the machine learning clustering method DBSCAN to distinguish between diffraction from multiple sources, and then calculates cluster size and inversion symmetry to assess whether clusters of intensity belong to desired data or alien signals. This approach can dramatically reduce the amount of time spent manually processing data, allowing BCDI data processing capabilities to keep pace with the technological advances of fourth-generation synchrotron light sources.

36 MATERIALS SCIENCE↗

Understanding the Existence of a Na 2 Dimer in a High-Spin State

The recent observation of a high-spin Na 2 dimer formed on the surface of liquid helium nanodroplets raises some fundamental questions, as the ground state of Na 2 is known to have zero spin. Is it protected against spontaneous dissociation? What is its binding energy and interatomic distance? Is it stable at a higher temperature? Using calculations based on density functional theory (with and without long-range interaction) and coupled cluster methods, CCSD(T), we show that the bonding in the high-spin Na 2 dimer is governed by van der Waals interaction with binding energy (bond length) varying between −0.030 eV (5.108 Å) and −0.192 eV (4.231 Å), depending on the computational method used. Thus, the experimental method used by Kresin and coworkers can be very useful to study larger metastable high-spin clusters such as Li 4 which was predicted in 1985 to have a tetrahedral structure carrying a magnetic moment of 2 μ B , while its ground state is planar and nonmagnetic.

Basis sets↗

Diffusion Monte Carlo evaluation of disiloxane linearisation barrier

The disiloxane molecule is a prime example of silicate compounds containing the Si-O-Si bridge. The molecule is of significant interest within the field of quantum chemistry, owing to the difficulty in theoretically predicting its properties. Herein, the linearisation barrier of disiloxane is investigated using a fixed-node diffusion Monte Carlo (FNDMC) approach, which is one of the most reliable ab initio methods in accounting for the electronic correlation. Calculations utilizing the density functional theory (DFT) and the coupled cluster method with single and double substitutions, including noniterative triples (CCSD(T)) are carried out alongside FNDMC for comparison. It is concluded that FNDMC successfully predicts the disiloxane linearisation barrier and does not depend on the completeness of the basis-set as much as DFT or CCSD(T), thus establishing its suitability.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Fock-space relativistic coupled-cluster calculations of two-valence atoms

We have developed an all-particle Fock-space relativistic coupled-cluster method for two-valence atomic systems. We then describe a scheme to employ the coupled-cluster wave function to calculate atomic properties. Based on these developments we calculate the excitation energies, magnetic hyperfine structure constants, and electric dipole matrix elements of Sr, Ba, and Yb. Furthermore, we calculate the electric quadrupole hyperfine structure constants and the electric dipole matrix elements of Sr{sup +}, Ba{sup +}, and Yb{sup +}. For these we use the one-valence coupled-cluster wave functions obtained as an intermediate in the two-valence calculations. We also calculate the magnetic dipole hyperfine structure constants of Yb{sup +}.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Feature Selection and Inferential Procedures for Video Data [Slides]

With the rise of electronic and high-dimensional data, new and innovative feature detection and statistical methods are required to perform accurate and meaningful statistical analysis of these datasets that provide unique statistical challenges. In the area of feature detection, much of the recent feature detection research in the computer vision community has focused on deep learning methods, which require large amounts of labeled training data. However, in many application areas, training data is very limited and often difficult to obtain. We develop methods for fast, unsupervised, precise feature detection for video data based on optical flows, edge detection, and clustering methods. We also use pretrained neural networks and interpretable linear models to extract features using very limited training data. In the area of statistics, while high-dimensional data analysis has been a main focus of recent statistical methodological research, much focus has been on populations of high-dimensional vectors, rather than populations of high-dimensional tensors, which are three-dimensional arrays that can be used to model dependent images, such as images taken of the same person or ripped video frames. Our feature detection method is a non-model-based method that fusses information from dense optical flow, raw image pixels, and frame differences to generate detections. Our hypothesis testing methods are based on the assumption that dependent images are concatenated into a tensor that follows a tensor normal distribution, and from this assumption, we derive likelihood-ratio, score, and regression-based tests for one- and multiple-sample testing problems. Our methods will be illustrated on simulated and real datasets. We conclude this report with comments on the relationship between feature detection and hypothesis testing methods.

97 MATHEMATICS AND COMPUTING↗

Label Assist: Personalized Travel Models for Longitudinal Data Collection

Understanding travel behavior is crucial to transportation decarbonization. OpenPATH is an open-source mobility platform which collects and analyzes human travel behavior at the individual level. The mobile application passively senses trips and prompts users to label them. However, users find the labeling process burdensome; less than half the trips are typically labeled, making much of the data unusable in aggregate analyses of mobility patterns. Prior work has addressed the response fatigue challenge through automated mode inference using sensor data, but sensors cannot capture all aspects of travel behavior. We explore an alternative approach in which we leverage prior user input to predict travel choices in novel trips. We first explore trip clustering methods and develop a novel two-step pipeline using DBSCAN and SVMs to extract realistic geospatial clusters. We then propose two strategies to predict trip labels: (i) clustering trips and extrapolating labels for similar trips, and (ii) random forest classification. The random forest approach is able to achieve - $70-80% accuracy (purpose: 72%, mode: 79%, replaced mode: 81%). These novel approaches to trip classification allow us to increase the rate of user labeling by suggesting predicted labels to be verified by the user. Unlabeled trips can also contribute to aggregate analyses, using label predictions and their associated confidences as a substitute. While there exist other travel survey apps with the ability to infer travel choices, to our knowledge, this is the first paper to describe such a supervised system and rigorously evaluate it.

ADVANCED PROPULSION SYSTEMS,ENERGY PLANNING, POLIC↗