Ti（C, N） multilayer films have been prepared by closed-field unbalanced magnetron sputtering technology and using graphite target as the C supplier. Microstructural observation results showed that the Ti（C, N） fil...Ti（C, N） multilayer films have been prepared by closed-field unbalanced magnetron sputtering technology and using graphite target as the C supplier. Microstructural observation results showed that the Ti（C, N） films exhibited multilayer structure with most of fine nano-columnar Ti（C, N） grains existing in the films. The current of graphite target had an effect remarkably on the multilayer structure of films： the periodical thickness gradually increased as the current went up, but the grain size of films gradually decreased and even amorphous phase appeared as the current further increased. The microstructure of Ti（C, N） films changed from columnar crystallite to nanocomposite in high current of graphite target where the fine Ti（C, N） grains were distributed uniformly in the amorphous Ti（C, N） matrix, and the volume fraction of the amorphous phase increased with increasing current. Measurement results showed that the Ti（C, N） multilayer films have high rnicrohardness and low friction coefficient, and especially the film deposited in the current of 0.9 A exhibits superior properties with optimizing hardness and friction coefficient. Based on the relationship of the microstructure and the properties of films, the multilayer structure and fine grain size of Ti（C, N） films are responsible for their well mechanical and friction properties. And choosing the graphite target as the C supplier is more propitious to decrease the friction coefficients of films.展开更多
This paper studies the generalization capability of feedforward neural networks (FNN).The mechanism of FNNs for classification is investigated from the geometric and probabilistic viewpoints. It is pointed out that th...This paper studies the generalization capability of feedforward neural networks (FNN).The mechanism of FNNs for classification is investigated from the geometric and probabilistic viewpoints. It is pointed out that the outputs of the output layer in the FNNs for classification correspond to the estimates of posteriori probability of the input pattern samples with desired outputs 1 or 0. The theorem for the generalized kernel function in the radial basis function networks (RBFN) is given. For an 2-layer perceptron network (2-LPN). an idea of using extended samples to improve generalization capability is proposed. Finally. the experimental results of radar target classification are given to verify the generaliztion capability of the RBFNs.展开更多
基金supported by the National Natural Science Foundation of China (No. 50971097)Shaanxi Provincial Project of Special Foundation of Key Disciplines
文摘Ti（C, N） multilayer films have been prepared by closed-field unbalanced magnetron sputtering technology and using graphite target as the C supplier. Microstructural observation results showed that the Ti（C, N） films exhibited multilayer structure with most of fine nano-columnar Ti（C, N） grains existing in the films. The current of graphite target had an effect remarkably on the multilayer structure of films： the periodical thickness gradually increased as the current went up, but the grain size of films gradually decreased and even amorphous phase appeared as the current further increased. The microstructure of Ti（C, N） films changed from columnar crystallite to nanocomposite in high current of graphite target where the fine Ti（C, N） grains were distributed uniformly in the amorphous Ti（C, N） matrix, and the volume fraction of the amorphous phase increased with increasing current. Measurement results showed that the Ti（C, N） multilayer films have high rnicrohardness and low friction coefficient, and especially the film deposited in the current of 0.9 A exhibits superior properties with optimizing hardness and friction coefficient. Based on the relationship of the microstructure and the properties of films, the multilayer structure and fine grain size of Ti（C, N） films are responsible for their well mechanical and friction properties. And choosing the graphite target as the C supplier is more propitious to decrease the friction coefficients of films.
文摘This paper studies the generalization capability of feedforward neural networks (FNN).The mechanism of FNNs for classification is investigated from the geometric and probabilistic viewpoints. It is pointed out that the outputs of the output layer in the FNNs for classification correspond to the estimates of posteriori probability of the input pattern samples with desired outputs 1 or 0. The theorem for the generalized kernel function in the radial basis function networks (RBFN) is given. For an 2-layer perceptron network (2-LPN). an idea of using extended samples to improve generalization capability is proposed. Finally. the experimental results of radar target classification are given to verify the generaliztion capability of the RBFNs.
