The transfer-integral (TI) method, developed especially for hard-core potentials, is tested with anx2 + 3x3 + 4x4 potential by a comparison with other methods: In the harmonic approximation the improvement of the TI method over the self-consistent phonon (SCP) method is about as great as that of the SCP method over the random phase approximation (RPA). 相似文献
We present a maximum margin parameter learning algorithm for Bayesian network classifiers using a conjugate gradient (CG) method for optimization. In contrast to previous approaches, we maintain the normalization constraints on the parameters of the Bayesian network during optimization, i.e., the probabilistic interpretation of the model is not lost. This enables us to handle missing features in discriminatively optimized Bayesian networks. In experiments, we compare the classification performance of maximum margin parameter learning to conditional likelihood and maximum likelihood learning approaches. Discriminative parameter learning significantly outperforms generative maximum likelihood estimation for naive Bayes and tree augmented naive Bayes structures on all considered data sets. Furthermore, maximizing the margin dominates the conditional likelihood approach in terms of classification performance in most cases. We provide results for a recently proposed maximum margin optimization approach based on convex relaxation. While the classification results are highly similar, our CG-based optimization is computationally up to orders of magnitude faster. Margin-optimized Bayesian network classifiers achieve classification performance comparable to support vector machines (SVMs) using fewer parameters. Moreover, we show that unanticipated missing feature values during classification can be easily processed by discriminatively optimized Bayesian network classifiers, a case where discriminative classifiers usually require mechanisms to complete unknown feature values in the data first. 相似文献
In diaphragm-based micromachined calorimetric flow sensors, convective heat transfer through the test fluid competes with the spurious heat shunt induced by the thin-film diaphragm where heating and temperature sensing elements are embedded. Consequently, accurate knowledge of thermal conductivity, thermal diffusivity, and emissivity of the diaphragm is mandatory for design, simulation, optimization, and characterization of such devices. However, these parameters can differ considerably from those stated for bulk material and they typically depend on the production process. We developed a novel technique to extract the thermal thin-film properties directly from measurements carried out on calorimetric flow sensors. Here, the heat transfer frequency response from the heater to the spatially separated temperature sensors is measured and compared to a theoretically obtained relationship arising from an extensive two-dimensional analytical model. The model covers the heat generation by the resistive heater, the heat conduction within the diaphragm, the radiation loss at the diaphragm’s surface, and the heat sink caused by the supporting silicon frame. This contribution summarizes the analytical heat transfer analysis in the microstructure and its verification by a computer numerical model, the measurement setup, and the associated thermal parameter extraction procedure. Furthermore, we report on measurement results for the thermal conductivity, thermal diffusivity, and effective emissivity obtained from calorimetric flow sensor specimens featuring dielectric thin-film diaphragms made of plasma enhanced chemical vapor deposition silicon nitride. 相似文献
During and after rolling or flattening of metal strips and plates the permissible deviations from flatness are described by the permissible absolute wave height and the flatness index. Both values can be determined on a measuring table while the material is not subjected to global tension. Because this procedure is expensive, time‐consuming and allows measurement only at discrete positions along the strip length, on‐line flatness measuring systems are used which can detect the distribution of longitudinal tensile stresses distributed across the strip width allowing for the calculation of the flatness index. This value does not always agree with the value obtained directly by measuring on the table even when the measurement of the longitudinal tensile stress distribution operates perfectly. It can be shown that the measurement of the tensile stress distribution does not give a direct indication on the wave height in the tension‐free state determined on the measuring table. To explain the relationship between tensile stress distribution in the strip and the flatness measurement on the measuring table, the buckling behavior is analysed both with and without dead load for simple symmetrical residual stress distributions resulting, e.g., from the rolling process. Based on the knowledge of the distribution of the longitudinal residual stresses across the strip width, the flatness index and the wave height can be determined by using a specialized finite element model. If the direct measurement is performed under action of dead load, large differences between the directly and indirectly obtained flatness index are observed. Below a certain limit of the intensity of the residual stress distribution the strips and plates lie flat on the measuring table. Above this limit the strip lying on the table exhibits post‐bucking deformations. In the latter case, the wave height increases with strip thickness and intensity of residual stresses. 相似文献
Traditionally, dynamic linkers merely combined previously compiled pieces of code. Faster processors are now making outright code generation at load time practical, leading to cross-platform portability at very little extra cost. While the central idea behind dynamic linking is quite straightforward, it can be implemented through a surprising variety of strategies. The author contrasts three simple dynamic linking schemes with two much more elaborate strategies. The latter represent a workload shift from the compiler to the dynamic linker, just as dynamic linking itself represents a shift that moves the functions of a separate linker into the loader. Because the new techniques promise the profound additional benefit of cross-platform portability, they will most likely displace the currently popular linking-loader approach 相似文献
Low-energy ion-assisted magnetron sputter deposition has been used for the synthesis of highly reflective Ni/V multilayer soft X-ray mirrors. A low ion energy and a high ion-to-metal flux ratio were employed in order to stimulate the adatom mobility while minimizing ion-induced intermixing at the interfaces. An analytic model, based on the binary collision approximation, was used in order to gain insight into low-energy ion–surface interactions as a function of ion energy and ion-to-metal flux ratio. The model predicted a favorable region in the ion energy-flux parameter space where only surface atomic displacements are stimulated during growth of Ni and V for multilayers. For a series of Ni/V multilayer mirrors with multilayer periods about Λ = 1.2 nm, grown with a continuous ion assistance using energies in the range 7–36 eV and with ion-to-metal flux ratios ΦNi = 4.7 and ΦV=20.9, specular and diffuse X-ray scattering analyses revealed that ion energies of 27–31 eV produced the best trade-off between reduced interfacial roughness and intermixing. However, it was also concluded that an interface mixing of about ± 1 atomic distance is unavoidable when a continuous flux of assisting ions is used.
To overcome this limitation, a sophisticated interface engineering technique was employed, where the first 0.3 nm of each layer was grown with a high-flux low-energy ion assistance and the remaining part was grown with a slightly higher ion energy. This method was demonstrated to largely eliminate the intermixing while maintaining the smoothening effect of ion assistance. Two Ni/V multilayer soft X-ray mirror structures, one with 500 periods designed for near-normal incidence and one 150 periods reflecting polarizer at the Brewster angle, were grown utilizing the interface engineering concept. Both the near-normal incidence reflectivity as well as polarizability were improved by a factor of 2 as compared to previously reported data for an X-ray energy of E = 511 eV. 相似文献