川南石刻图像特征提取方法研究

本文以四川南部地区南宋墓葬石刻艺术为切入点,提出石刻图像特征提取的具体方法,并通过仿真试验证明了该方法的有效性,最后指出了进一步的研究方向.     

Redundant multicast routing in multilayer networks with shared risk resource groups: Complexity,models and algorithms

This paper presents a redundant multicast routing problem in multilayer networks that arises from large-scale distribution of realtime multicast data (e.g., Internet TV, videocasting, online games, stock quotes). Since these multicast services commonly operate in multilayer networks, the communications paths need to be robust against a single router or link failure as well as multiple such failures due to shared risk link groups (SRLGs). The main challenge of this multicast is to ensure the service availability and reliability using a path protection scheme, which is to find a redundant path that is SRLG-disjoint (diverse) from each working path. The objective of this problem is, therefore, to find two redundant multicast trees, each from one of the two redundant sources to every destination, at a minimum total communication cost whereas two paths from the two sources to every destination are guaranteed to be SRLG-diverse (i.e., links in the same risk group are disjoint). In this paper, we present two new mathematical programming models, edge-based and path-based, for the redundant multicast routing problem with SRLG-diverse constraints. Because the number of paths in path-based model grows exponentially with the network size, it is impossible to enumerate all possible paths in real life networks. We develop three approaches (probabilistic, non-dominated and nearly non-dominated) to generate potentially good paths that may be included in the path-based model. This study is motivated by emerging applications of internet-protocol TV service, and we evaluate the proposed approaches using real life network topologies. Our empirical results suggest that both models perform very well, and the nearly non-dominated path approach outperforms all other path generation approaches.     

Light-induced electron spin polarization in the ground state of water-soluble copper porphyrins

Light-induced spin-polarized transient EPR spectra are reported for several water-soluble copper porphyrins. The spectra are assigned to the doublet ground state, with emissive spin polarization resulting from photoexcitation and subsequent electronic relaxation. In contrast to other systems for which polarization of a doublet ground state has been observed, the exchange interactions in the copper porphyrins are strong and the geometry is fixed. It is proposed that intersystem crossing from the photoexcited trip-doublet to the trip-quartet state can lead to net polarization of the spin system and that this polarization is maintained during electronic decay, possibly via charge-transfer and exciplex states. The intensity of the observed spin polarization is essentially independent of the molecular orientation in the external field, but is strongly dependent on the nature of the charged peripheral groups. Possible reasons for this behavior are discussed.     

Individual Particle Morphology and Acidity

The morphological characterization of particles during the Pittsburgh Air Quality Study (PAQS) suggests that particle shape and physical state depends on their acidity. The aerosol shape parameters measured by Computer-Controlled Scanning Electron Microscopy (CCSEM) are statistically different in periods when atmospheric particles are neutral and when they are acidic. High concentrations of particles smaller than 500 nm with high sulfur content and liquid appearance or signs of a previous liquid state before partial or total recrystallization are present on filters collected in days with high aerosol acidity. By contrast, in winter and summer periods in which the aerosol was neutral, the shape parameter values are similar and suggest that the particles have been dried out. These direct observations support the hypothesis that during summer acidic periods in the Eastern United States the particles may contain some water even at low relative humidity both in the atmosphere and on filters.     

数字时代的设计观念与技术

对数字时代的设计技术进行了研究，探讨了现代设计观念。     

The effect of the third invariant on the strength and failure of boron carbide and silicon carbide

This article presents new test data to assess the effect the third invariant has on the strength and failure of two ceramic materials: boron carbide and silicon carbide. Two experimental techniques are used: the Brazilian test that produces a biaxial state of stress, and a new technique that uses a high-pressure confinement vessel to load a specially designed dumbbell specimen in triaxial extension. The dumbbell geometry provides two important advantages over the typically used cylindrical specimen: no adhesive is required to bond the specimen to the load cell because the dumbbell geometry naturally takes the specimen into tension, and any loading asymmetries are essentially eliminated due to the axisymmetric geometry. The results show that when the stress state is on the tensile meridian the equivalent stress at failure is constant, independent of the hydrostatic pressure. The average equivalent stress at failure is for boron carbide and for silicon carbide. The Brazilian test was only performed on boron carbide and failed at , much higher than when on the tensile meridian () indicating that the effect of the third invariant is significant (because of the difference in the failure strength) and must be accounted for to accurately predict when failure will occur.     

斯巴克 Cayin A-300B综合功放

即使是那些短视的批评家们现在也不得不承认,自十年前起重现美国国内音频市场的单端真空管功放,不仅仅代表了一股逝去的潮流.尽管这类同任何一款普通扬声器搭配都有问题的功放都决不会成为司空见惯的产品,但同为现实的是,SET(单端真空管功放)的魅力已让它在市场中永远占据着一席之地.总会有制造商想要生产SET,正如总有人想要扮演莎士比亚或蒸馏苦艾酒.     

Maximization of mutual information for supervised linear feature extraction.

In this paper, we present a novel scheme for linear feature extraction in classification. The method is based on the maximization of the mutual information (MI) between the features extracted and the classes. The sum of the MI corresponding to each of the features is taken as an heuristic that approximates the MI of the whole output vector. Then, a component-by-component gradient-ascent method is proposed for the maximization of the MI, similar to the gradient-based entropy optimization used in independent component analysis (ICA). The simulation results show that not only is the method competitive when compared to existing supervised feature extraction methods in all cases studied, but it also remarkably outperform them when the data are characterized by strongly nonlinear boundaries between classes.