Approximation of hyperbolic tangent activation function using hybrid methods

Artificial Neural Networks are widely used in various applications in engineering, as such solutions of nonlinear problems. The implementation of this technique in reconfigurable devices is a great challenge to researchers by several factors, such as floating point precision, nonlinear activation function, performance and area used in FPGA. The contribution of this work is the approximation of a nonlinear function used in ANN, the popular hyperbolic tangent activation function. The system architecture is composed of several scenarios that provide a tradeoff of performance, precision and area used in FPGA. The results are compared in different scenarios and with current literature on error analysis, area and system performance. © 2013 IEEE.

Pruning methods to MLP neural networks considering proportional apparent error rate for classification problems with unbalanced data

This article deals with classification problems involving unequal probabilities in each class and discusses metrics to systems that use multilayer perceptrons neural networks (MLP) for the task of classifying new patterns. In addition we propose three new pruning methods that were compared to other seven existing methods in the literature for MLP networks. All pruning algorithms presented in this paper have been modified by the authors to do pruning of neurons, in order to produce fully connected MLP networks but being small in its intermediary layer. Experiments were carried out involving the E. coli unbalanced classification problem and ten pruning methods. The proposed methods had obtained good results, actually, better results than another pruning methods previously defined at the MLP neural network area. (C) 2014 Elsevier Ltd. All rights reserved.

Hybrid density functional study of small Rh-n (n=2-15) clusters

The physical properties of small rhodium clusters, Rh-n, have been in debate due to the shortcomings of density functional theory (DFT). To help in the solution of those problems, we obtained a set of putative lowest energy structures for small Rh-n (n = 2-15) clusters employing hybrid-DFT and the generalized gradient approximation (GGA). For n = 2-6, both hybrid and GGA functionals yield similar ground-state structures (compact), however, hybrid favors compact structures for n = 7-15, while GGA favors open structures based on simple cubic motifs. Thus, experimental results are crucial to indicate the correct ground-state structures, however, we found that a unique set of structures (compact or open) is unable to explain all available experimental data. For example, the GGA structures (open) yield total magnetic moments in excellent agreement with experimental data, while hybrid structures (compact) have larger magnetic moments compared with experiments due to the increased localization of the 4d states. Thus, we would conclude that GGA provides a better description of the Rh-n clusters, however, a recent experimental-theoretical study [ Harding et al., J. Chem. Phys. 133, 214304 (2010)] found that only compact structures are able to explain experimental vibrational data, while open structures cannot. Therefore, it indicates that the study of Rh-n clusters is a challenging problem and further experimental studies are required to help in the solution of this conundrum, as well as a better description of the exchange and correlation effects on the Rh n clusters using theoretical methods such as the quantum Monte Carlo method.

Saturated flow boiling heat transfer and critical heat flux in small horizontal flattened tubes

This paper presents experimental results for flow boiling heat transfer coefficient and critical heat flux (CHF) in small flattened tubes. The tested flattened tubes have the same equivalent internal diameter of 2.2 mm, but different aspect height/width ratios (H/W) of 1/4, 1/2, 2 and 4. The experimental data were compared against results for circular tubes using R134a and R245fa as working fluids at a nominal saturation temperature of 31 degrees C. For mass velocities higher than 200 kg/m(2)s, the flattened and circular tubes presented similar heat transfer coefficients. Such a behavior is related to the fact that stratification effects are negligible under conditions of higher mass velocities. Heat transfer correlations from the literature, usually developed using only circular-channel experimental data, predicted the flattened tube results for mass velocities higher than 200 kg/m(2)s with mean absolute error lower than 20% using the equivalent diameter to account for the geometry effect. Similarly, the critical heat flux results were found to be independent of the tube aspect ratio when the same equivalent length was kept. Equivalent length is a new parameter which takes into account the channel heat transfer area. The CHF correlations for round tubes predicted the flattened tube data relatively well when using the equivalent diameter and length. Furthermore, a new proposed CHF correlation predicted the present flattened tube data with a mean absolute error of 5%. (C) 2012 Elsevier Ltd. All rights reserved.

Quantum Monte Carlo study of small aluminum clusters Al-n (n=2-13)

Using fixed node diffusion quantum Monte Carlo (FN-DMC) simulations and density functional theory (DFT) within the generalized gradient approximations, we calculate the total energies of the relaxed and unrelaxed neutral, cationic, and anionic aluminum clusters, Al-n (n = 1-13). From the obtained total energies, we extract the ionization potential and electron detachment energy and compare with previous theoretical and experimental results. Our results for the electronic properties from both the FN-DMC and DFT calculations are in reasonably good agreement with the available experimental data. A comparison between the FN-DMC and DFT results reveals that their differences are a few tenths of electron volt for both the ionization potential and the electron detachment energy. We also observe two distinct behaviors in the electron correlation contribution to the total energies from smaller to larger clusters, which could be assigned to the structural transition of the clusters from planar to three-dimensional occurring at n = 4 to 5.

Encapsulation of small magnetic clusters in fullerene cages: A density functional theory investigation within van der Waals corrections

The encapsulation of magnetic transition-metal (TM) clusters inside carbon cages (fullerenes, nanotubes) has been of great interest due to the wide range of applications, which spread from medical sensors in magnetic resonance imaging to photonic crystals. Several theoretical studies have been reported; however, our atomistic understanding of the physical properties of encapsulated magnetic TM 3d clusters is far from satisfactory. In this work, we will report general trends, derived from density functional theory within the generalized gradient approximation proposed by Perdew, Burke, and Ernzerhof (PBE), for the encapsulation properties of the TMm@C-n (TM = Fe, Co, Ni; m = 2-6, n = 60,70,80,90) systems. Furthermore, to understand the role of the van der Waals corrections to the physical properties, we employed the empirical Grimme's correction (PBE + D2). We found that both PBE and PBE + D2 functionals yield almost the same geometric parameters, magnetic and electronic properties, however, PBE + D2 strongly enhances the encapsulation energy. We found that the center of mass of the TMm clusters is displaced towards the inside C-n surfaces, except for large TMm clusters (m = 5 and 6). For few cases, e. g., Co-4 and Fe-4, the encapsulation changes the putative lowest-energy structure compared to the isolated TMm clusters. We identified few physical parameters that play an important role in the sign and magnitude of the encapsulation energy, namely, cluster size, fullerene equatorial diameter, shape, curvature of the inside C-n surface, number of TM atoms that bind directly to the inside C-n surface, and the van der Waals correction. The total magnetic moment of encapsulated TMm clusters decreases compared with the isolated TMm clusters, which is expected due to the hybridization of the d-p states, and strongly depends on the size and shape of the fullerene cages.

Hochgenaue Berechnungen von Energien und molekularen Eigenschaften mit Hilfe der Coupled-Cluster-Theorie

Coupled-Cluster-Theorie (CC) ist in der heutigen Quantenchemie eine der erfolgreichsten Methoden zur genauen Beschreibung von Molekülen. Die in dieser Arbeit vorgestellten Ergebnisse zeigen, daß neben den Berechnungen von Energien eine Reihe von Eigenschaften wie Strukturparameter, Schwingungsfrequenzen und Rotations-Schwingungs-Parameter kleiner und mittelgrofler Moleküle zuverlässig und präzise vorhergesagt werden können. Im ersten Teil der Arbeit wird mit dem Spin-adaptierten Coupled-Cluster-Ansatz (SA-CC) ein neuer Weg zur Verbesserung der Beschreibung von offenschaligen Systemen vorgestellt. Dabei werden zur Bestimmung der unbekannten Wellenfunktionsparameter zusätzlich die CC-Spingleichungen gelöst. Durch dieses Vorgehen wird gewährleistet, daß die erhaltene Wellenfunktion eine Spineigenfunktion ist. Die durchgeführte Implementierung des Spin-adaptierten CC-Ansatzes unter Berücksichtigung von Einfach- und Zweifachanregungen (CCSD) für high-spin Triplett-Systeme wird ausführlich erläutert. Im zweiten Teil werden CC-Additionsschemata vorgestellt, die auf der Annahme der Additivität von Elektronenkorrelations- und Basissatzeffekten basieren. Die etablierte Vorgehensweise, verschiedene Beiträge zur Energie mit an den Rechenaufwand angepassten Basissätzen separat zu berechnen und aufzusummieren, wird hier auf Gradienten und Kraftkonstanten übertragen. Für eine Beschreibung von Bindungslängen und harmonischen Schwingungsfrequenzen mit experimenteller Genauigkeit ist die Berücksichtigung von Innerschalenkorrelationseffekten sowie Dreifach- und Vierfachanregungen im Clusteroperator der Wellenfunktion nötig. Die Basissatzkonvergenz wird dabei zusätzlich mit Extrapolationsmethoden beschleunigt. Die quantitative Vorhersage der Bindungslängen von 17 kleinen Molekülen, aufgebaut aus Atomen der ersten Langperiode, ist so mit einer Genauigkeit von wenigen Hundertstel Pikometern möglich. Für die Schwingungsfrequenzen dieser Moleküle weist das CC-Additionsschema basierend auf den berechneten Kraftkonstanten im Vergleich zu experimentellen Ergebnissen einen mittleren absoluten Fehler von 3.5 cm-1 und eine Standardabweichung von 2.2 cm-1 auf. Darüber hinaus werden zur Unterstützung von experimentellen Untersuchungen berechnete spektroskopische Daten einiger größerer Moleküle vorgelegt. Die in dieser Arbeit vorgestellten Untersuchungen zur Isomerisierung von Dihalogensulfanen XSSX (X= F, Cl) oder die Berechnung von Struktur- und Rotations-Schwingungs-Parametern für die Moleküle CHCl2F und CHClF2 zeigen, daß bereits störungstheoretische CCSD(T)-Näherungsmethoden qualitativ gute Vorhersagen experimenteller Resultate liefern. Desweiteren werden Diskrepanzen von experimentellen und berechneten Bindungsabständen bei den Molekülen Borhydrid- und Carbenylium durch die Berücksichtigung des elektronischen Beitrages zum Trägheitsmoment beseitigt.

Übergangsmomente zwischen angeregten Zuständen mit der RI-CC2-Methode: Implementierung und Anwendung auf Triplett-Excimere

Die Themengebiete dieser Arbeit umfassen sowohl methodische Weiterentwicklungen im Rahmen der ab initio zweiter Ordnungsmethoden CC2 und ADC(2) als auch Anwendungen dieser Weiterentwick-lungen auf aktuelle Fragestellungen. Die methodischen Erweiterungen stehen dabei hauptsächlich im Zusammenhang mit Übergangsmomenten zwischen angeregten Zuständen. Durch die Implementie-rung der selbigen ist nun die Berechnung transienter Absorptionsspektren möglich. Die Anwendungen behandeln vorwiegend das Feld der organischen Halbleiter und deren photo-elektronische Eigen-schaften. Dabei spielen die bislang wenig erforschten Triplett-Excimere eine zentrale Rolle.rnDie Übergangsmomente zwischen angeregten Zuständen wurden in das Programmpaket TUR-BOMOLE implementiert. Dadurch wurde die Berechnung der Übergangsmomente zwischen Zustän-den gleicher Multiplizität (d.h. sowohl Singulett-Singulett- als auch Triplett-Triplett-Übergänge) und unterschiedlicher Multiplizität (also Singulett-Triplett-Übergänge) möglich. Als Erweiterung wurde durch ein Interface zum ORCA Programm die Berechnung von Spin-Orbit-Matrixelementen (SOMEs) implementiert. Des Weiteren kann man mit dieser Implementierung auch Übergänge in offenschaligen Systemen berechnen. Um den Speicherbedarf und die Rechenzeit möglichst gering zu halten wurde die resolution-of-the-identity (RI-) Näherung benutzt. Damit lässt sich der Speicherbedarf von O(N4) auf O(N3) reduzieren, da die mit O(N4) skalierenden Größen (z. B. die T2-Amplituden) sehr effizient aus RI-Intermediaten berechnet werden können und daher nicht abgespeichert werden müssen. Dadurch wird eine Berechnung für mittelgroße Moleküle (ca. 20-50 Atome) mit einer angemessenen Basis möglich.rnDie Genauigkeit der Übergangsmomente zwischen angeregten Zuständen wurde für einen Testsatz kleiner Moleküle sowie für ausgewählte größere organische Moleküle getestet. Dabei stellte sich her-aus, dass der Fehler der RI-Näherung sehr klein ist. Die Vorhersage der transienten Spektren mit CC2 bzw. ADC(2) birgt allerdings ein Problem, da diese Methoden solche Zustände nur sehr unzureichend beschreiben, welche hauptsächlich durch zweifach-Anregungen bezüglich der Referenzdeterminante erzeugt werden. Dies ist für die Spektren aus dem angeregten Zustand relevant, da Übergänge zu diesen Zuständen energetisch zugänglich und erlaubt sein können. Ein Beispiel dafür wird anhand eines Singulett-Singulett-Spektrums in der vorliegenden Arbeit diskutiert. Für die Übergänge zwischen Triplettzuständen ist dies allerdings weniger problematisch, da die energetisch niedrigsten Doppelan-regungen geschlossenschalig sind und daher für Tripletts nicht auftreten.rnVon besonderem Interesse für diese Arbeit ist die Bildung von Excimeren im angeregten Triplettzu-stand. Diese können aufgrund starker Wechselwirkungen zwischen den π-Elektronensystemen großer organischer Moleküle auftreten, wie sie zum Beispiel als organische Halbleiter in organischen Leucht-dioden eingesetzt werden. Dabei können die Excimere die photo-elktronischen Eigenschaften dieser Substanzen signifikant beeinflussen. Im Rahmen dieser Dissertation wurden daher zwei solcher Sys-teme untersucht, [3.3](4,4’)Biphenylophan und das Naphthalin-Dimer. Hierzu wurden die transienten Anregungsspektren aus dem ersten angeregten Triplettzustand berechnet und diese Ergebnisse für die Interpretation der experimentellen Spektren herangezogen. Aufgrund der guten Übereinstimmung zwischen den berechneten und den experimentellen Spektren konnte gezeigt werden, dass es für eine koplanare Anordnung der beiden Monomere zu einer starken Kopplung zwischen lokal angereg-ten und charge-transfer Zuständen kommt. Diese Kopplung resultiert in einer signifikanten energeti-schen Absenkung des ersten angeregten Zustandes und zu einem sehr geringen Abstand zwischen den Monomereinheiten. Dabei ist der angeregte Zustand über beide Monomere delokalisiert. Die star-ke Kopplung tritt bei einem intermolekularen Abstand ≤4 Å auf, was einem typischen Abstand in orga-nischen Halbleitern entspricht. In diesem Bereich kann man zur Berechnung dieser Systeme nicht auf die Förster-Dexter-Theorie zurückgreifen, da diese nur für den Grenzfall der schwachen Kopplung gültig ist.

Benchmark Structures and Binding Energies of Small Water Clusters with Anharmonicity Corrections

For (H2O)n where n = 1–10, we used a scheme combining molecular dynamics sampling with high level ab initio calculations to locate the global and many low lying local minima for each cluster. For each isomer, we extrapolated the RI-MP2 energies to their complete basis set limit, included a CCSD(T) correction using a smaller basis set and added finite temperature corrections within the rigid-rotor-harmonic-oscillator (RRHO) model using scaled and unscaled harmonic vibrational frequencies. The vibrational scaling factors were determined specifically for water clusters by comparing harmonic frequencies with VPT2 fundamental frequencies. We find the CCSD(T) correction to the RI-MP2 binding energy to be small (<1%) but still important in determining accurate conformational energies. Anharmonic corrections are found to be non-negligble; they do not alter the energetic ordering of isomers, but they do lower the free energies of formation of the water clusters by as much as 4 kcal/mol at 298.15 K.

A modified test for small-study effects in meta-analyses of controlled trials with binary endpoints

Publication bias and related bias in meta-analysis is often examined by visually checking for asymmetry in funnel plots of treatment effect against its standard error. Formal statistical tests of funnel plot asymmetry have been proposed, but when applied to binary outcome data these can give false-positive rates that are higher than the nominal level in some situations (large treatment effects, or few events per trial, or all trials of similar sizes). We develop a modified linear regression test for funnel plot asymmetry based on the efficient score and its variance, Fisher's information. The performance of this test is compared to the other proposed tests in simulation analyses based on the characteristics of published controlled trials. When there is little or no between-trial heterogeneity, this modified test has a false-positive rate close to the nominal level while maintaining similar power to the original linear regression test ('Egger' test). When the degree of between-trial heterogeneity is large, none of the tests that have been proposed has uniformly good properties.

Inference on Survival Data with Covariate Measurement Error - An Imputation-based Approach

We propose a new method for fitting proportional hazards models with error-prone covariates. Regression coefficients are estimated by solving an estimating equation that is the average of the partial likelihood scores based on imputed true covariates. For the purpose of imputation, a linear spline model is assumed on the baseline hazard. We discuss consistency and asymptotic normality of the resulting estimators, and propose a stochastic approximation scheme to obtain the estimates. The algorithm is easy to implement, and reduces to the ordinary Cox partial likelihood approach when the measurement error has a degenerative distribution. Simulations indicate high efficiency and robustness. We consider the special case where error-prone replicates are available on the unobserved true covariates. As expected, increasing the number of replicate for the unobserved covariates increases efficiency and reduces bias. We illustrate the practical utility of the proposed method with an Eastern Cooperative Oncology Group clinical trial where a genetic marker, c-myc expression level, is subject to measurement error.

Control design and genetic algorithm optimization for electrostatic MEMS

This dissertation discusses structural-electrostatic modeling techniques, genetic algorithm based optimization and control design for electrostatic micro devices. First, an alternative modeling technique, the interpolated force model, for electrostatic micro devices is discussed. The method provides improved computational efficiency relative to a benchmark model, as well as improved accuracy for irregular electrode configurations relative to a common approximate model, the parallel plate approximation model. For the configuration most similar to two parallel plates, expected to be the best case scenario for the approximate model, both the parallel plate approximation model and the interpolated force model maintained less than 2.2% error in static deflection compared to the benchmark model. For the configuration expected to be the worst case scenario for the parallel plate approximation model, the interpolated force model maintained less than 2.9% error in static deflection while the parallel plate approximation model is incapable of handling the configuration. Second, genetic algorithm based optimization is shown to improve the design of an electrostatic micro sensor. The design space is enlarged from published design spaces to include the configuration of both sensing and actuation electrodes, material distribution, actuation voltage and other geometric dimensions. For a small population, the design was improved by approximately a factor of 6 over 15 generations to a fitness value of 3.2 fF. For a larger population seeded with the best configurations of the previous optimization, the design was improved by another 7% in 5 generations to a fitness value of 3.0 fF. Third, a learning control algorithm is presented that reduces the closing time of a radiofrequency microelectromechanical systems switch by minimizing bounce while maintaining robustness to fabrication variability. Electrostatic actuation of the plate causes pull-in with high impact velocities, which are difficult to control due to parameter variations from part to part. A single degree-of-freedom model was utilized to design a learning control algorithm that shapes the actuation voltage based on the open/closed state of the switch. Experiments on 3 test switches show that after 5-10 iterations, the learning algorithm lands the switch with an impact velocity not exceeding 0.2 m/s, eliminating bounce.

OPTIMAL SHAPE IN ELECTROMAGNETIC SCATTERING BY SMALL ASPHERICAL PARTICLES

We consider the question of optimal shapes, e.g., those causing minimal extinction among all shapes of equal volume. Guided by the isoperimetric property of a sphere, relevant in the geometrical optics limit of scattering by large particles, we examine an analogous question in the low frequency (electrostatics) approximation, seeking to disentangle electric and geometric contributions. To that end, we survey the literature on shape functionals and focus on ellipsoids, giving a simple proof of spherical optimality for the coated ellipsoidal particle. Monotonic increase with asphericity in the low frequency regime for orientation-averaged induced dipole moments and scattering cross-sections is also shown. Additional physical insight is obtained from the Rayleigh-Gans (transparent) limit and eccentricity expansions. We propose connecting low and high frequency regime in a single minimum principle valid for all size parameters, provided that reasonable size distributions of randomly oriented aspherical particles wash out the resonances for intermediate size parameters. This proposal is further supported by the sum rule for integrated extinction.

Discontinuous Galerkin finite element approximation of quasilinear elliptic boundary value problems II: Strongly monotone quasi-Newtonian flows

In this article, we develop the a priori and a posteriori error analysis of hp-version interior penalty discontinuous Galerkin finite element methods for strongly monotone quasi-Newtonian fluid flows in a bounded Lipschitz domain Ω ⊂ ℝd, d = 2, 3. In the latter case, computable upper and lower bounds on the error are derived in terms of a natural energy norm, which are explicit in the local mesh size and local polynomial degree of the approximating finite element method. A series of numerical experiments illustrate the performance of the proposed a posteriori error indicators within an automatic hp-adaptive refinement algorithm.

Multinomial goodness-of-fit: large sample tests with survey design correction and exact tests for small samples

I introduce the new mgof command to compute distributional tests for discrete (categorical, multinomial) variables. The command supports largesample tests for complex survey designs and exact tests for small samples as well as classic large-sample x2-approximation tests based on Pearson’s X2, the likelihood ratio, or any other statistic from the power-divergence family (Cressie and Read, 1984, Journal of the Royal Statistical Society, Series B (Methodological) 46: 440–464). The complex survey correction is based on the approach by Rao and Scott (1981, Journal of the American Statistical Association 76: 221–230) and parallels the survey design correction used for independence tests in svy: tabulate. mgof computes the exact tests by using Monte Carlo methods or exhaustive enumeration. mgof also provides an exact one-sample Kolmogorov–Smirnov test for discrete data.