Variational Bayes solution of linear neural networks and its generalization performance

It is well known that in unidentifiable models, the Bayes estimation provides much better generalization performance than the maximum likelihood (ML) estimation. However, its accurate approximation by Markov chain Monte Carlo methods requires huge computational costs. As an alternative, a tractable approximation method, called the variational Bayes (VB) approach, has recently been proposed and has been attracting attention. Its advantage over the expectation maximization (EM) algorithm, often used for realizing the ML estimation, has been experimentally shown in many applications; nevertheless, it has not yet been theoretically shown. In this letter, through analysis of the simplest unidentifiable models, we theoretically show some properties of the VB approach. We first prove that in three-layer linear neural networks, the VB approach is asymptotically equivalent to a positive-part James-Stein type shrinkage estimation. Then we theoretically clarify its free energy, generalization error, and training error. Comparing them with those of the ML estimation and the Bayes estimation, we discuss the advantage of the VB approach. We also show that unlike in the Bayes estimation, the free energy and the generalization error are less simply related with each other and that in typical cases, the VB free energy well approximates the Bayes one, while the VB generalization error significantly differs from the Bayes one.     

Multiobjective evolutionary computation for supersonic wing-shapeoptimization

This paper discusses the design optimization of a wing for supersonic transport (SST) using a multiple-objective genetic algorithm (MOGA). Three objective functions are used to minimize the drag for supersonic cruise, the drag for transonic cruise, and the bending moment at the wing root for supersonic cruise. The wing shape is defined by 66 design variables. A Euler flow code is used to evaluate supersonic performance, and a potential flow code is used to evaluate transonic performance. To reduce the total computational time, flow calculations are parallelized on an NEC SX-4 computer using 32 processing elements. The detailed analysis of the resulting Pareto front suggests a renewed interest in the arrow wing planform for the supersonic wing     

A pilot survey of 39 Victorian WWTP effluents using a high speed luminescent umu test in conjunction with a novel GC-MS-database technique for automatic identification of micropollutants

In 2007, samples of treated effluent were collected at point of discharge to the environment from 39 wastewater treatment plants (WWTPs) located across Victoria, Australia grouped by treatment type. Sample genotoxicity was assessed with a high-throughput luminescent umu test method using Salmonella typhimurium TL210 strain, with and without addition of a commercially available metabolic activation system. Samples were also screened using a gas chromatographic-mass spectrometric mass-structure database recognition method. A genotoxic response was observed in half of the samples tested without metabolic activation system (相似文献   

Loss mechanisms and high power piezoelectrics

Heat generation is one of the significant problems in piezoelectrics for high power density applications. In this paper, we review the loss mechanisms in piezoelectrics first, followed by the heat generation processes for various drive conditions. Heat generation at off-resonance is caused mainly by dielectric loss tan δ′ (i.e., P-E hysteresis loss), not by mechanical loss, while the heat generation at resonance is mainly attributed to mechanical loss tan φ′. Then, practical high power materials developed at Penn State is introduced, which exhibit the vibration velocity more than 1 m/s, leading to the power density capability 10 times of the commercially available “hard” PZTs. We propose a internal bias field model to explain the low loss and high power origin of these materials. Finally, using a low temperature sinterable “hard” PZT, we demonstrated a high power multilayer piezoelectric transformers.     

Sol–Gel‐Derived High‐Performance Stacked Transparent Conductive Oxide Thin Films

Single‐ and multi‐layer transparent conductive oxide (TCO) thin films exhibiting high performance, good packing density and low surface/interface roughness are deposited on silica glass substrates by the sol–gel method. The crystal and microstructural properties of the TCO thin films are evaluated as an alternate to films prepared by ultra‐high vacuum deposition. Tin‐doped indium oxide (ITO) thin films produced using a two‐step drying process showed low surface roughness because of dense packing structure not only horizontal but also vertical directions. As a result, electrical conductivity, carrier concentration, carrier mobility, and optical transmittance of 2.3 × 10³ S/cm, 8 × 10²⁰ cm^?3, 18 cm²/Vs, and over 98% at 500 nm, respectively, were achieved. A multilayer ZnO/ITO stacked structure was also fabricated using the sol–gel process. Our findings suggest that solution‐based methods show promise as an alternative to existing ultra‐high vacuum methods to fabricate TCO thin films.     

Micro machining for control of wettability with surface topography

A micro fabrication is presented to manufacture hydrophobic surfaces with micro-scale structures. Hydrophobicity is controlled with the shape and the alignment of micro pillars in the structure. The structures are manufactured in large areas at high production rates in the following processes: (1) the structure is fabricated on a tool by focused ion beam sputtering; (2) the reverse structure is formed on a metal plate by incremental stamping using the structured tool; and (3) the structure is transferred onto plastic plates by molding. A consecutive stamping is also proposed to fabricate several structures on a surface accurately with a structured tool, in which the moving pitch of the structured tool is numerically controlled. The effect of the surface topography on hydrophobicity is discussed with measuring contact angles on the structured surfaces in the water droplet tests. Hydrophobicity on the plastic plate is associated with the solid fraction on the structured surface based on the Cassie–Baxter model. A larger contact angle is observed for a smaller solid fraction of the surface.     

Transverse surface waves in a piezoelectric material carrying a gradient metal layer of finite thickness

The existence and propagation behavior of transverse surface waves in a layered structure concerning a piezoelectric substrate and a gradient metal layer are theoretically investigated in this paper. The Wentzel-Kramers-Brillouin (WKB) method is applied to obtain the analytical solutions in the gradient metal layer. The dispersion equation for transverse surface waves in such structure is obtained in a quite simple mathematical form, where the material gradient of the metal layer assumes arbitrary functions. Effects of material gradient on three types of dispersion behavior are discussed in detail based on a proper classification. Numerical results show that the material gradient in the metal layer evidently affects the fundamental mode shape of the transverse surface waves but has negligible effects on the higher order modes.     

Modeling and energy simulation of the variable refrigerant flow air conditioning system with water-cooled condenser under cooling conditions

As a new system, variable refrigerant flow system with water-cooled condenser (water-cooled VRF) can offer several interesting characteristics for potential users. However, at present, its dynamic simulation simultaneously in association with building and other equipments is not yet included in the energy simulation programs. Based on the EnergyPlus's codes, and using manufacturer's performance parameters and data, the special simulation module for water-cooled VRF is developed and embedded in the software of EnergyPlus. After modeling and testing the new module, on the basis of a typical office building in Shanghai with water-cooled VRF system, the monthly and seasonal cooling energy consumption and the breakdown of the total power consumption are analyzed. The simulation results show that, during the whole cooling period, the fan-coil plus fresh air (FPFA) system consumes about 20% more power than the water-cooled VRF system does. The power comparison between the water-cooled VRF system and the air-cooled VRF system is performed too. All of these can provide designers some ideas to analyze the energy features of this new system and then to determine a better scheme of the air conditioning system.     

Asymmetry in object substitution masking occurs relative to the direction of spatial attention shift.

A sparse mask that persists beyond the duration of a target can reduce its visibility, a phenomenon called object substitution masking. Y. Jiang and M. M. Chun (2001a) found an asymmetric pattern of substitution masking such that a mask on the peripheral side of the target caused stronger substitution masking than on the central side. Assuming spatial attention was focused toward the target, the peripheral and central masks were located in the same and opposite direction of an attentional path with reference to the target in their study. We hypothesized that this asymmetric mask configuration relative to the attentional shift contributes to asymmetric substitution masking. To test this hypothesis, we conducted four experiments among which the presence or absence of the center–periphery relationship and the presence or absence of the asymmetric mask configuration were manipulated independently and orthogonally. The results suggest that asymmetric substitution masking occurs relative to the direction of spatial attention irrespective of the central–peripheral relation. We propose that the asymmetry in substitution masking might be explained by attentional momentum associated with orienting toward the target. (PsycINFO Database Record (c) 2010 APA, all rights reserved)