The kernel and its parameters in support vector machine are important, an effect method for determining the parameter of Gaussian kernel based on the distances among the support vectors is proposed. The characters that the optimal discriminative function is determined by the support vectors, and the support vectors are centered as the Gaussian function, are considered in the method. Experimental results show that the method exhibits the essence of image feature space and solves a difficult problem for the parameter of Gaussian kernel in application.
To avoid the shortcomings in security, reliability and negative influence on original network performance, this paper brings forward a dynamic reconfiguration method based on information stream. Network dynamic reconfiguration implemented by this method can guarantee the important information can arrive at destination, and this method makes the information stream go through different ways, ensuring the whole performance of network to a considerable extent.
Traditional GIS-based oil spill information system can be used only on networked PCs which are physically tied to locations with limited accesses and restricted to certain platforms of operating systems. In order to provide a more convenient, accessible GIS application and response information to support timely decision in oil spill emergency response, the paper proposed the construction of marine oil spill response information system based on WebGIS. Based on the system structure and functions, the authors developed the oil spill behavior simulation module using numerical modeling technique, established the response information databases, developed the system server and client application programs by using GIS, dynamic Web, database techniques, and also discussed the system integration and visual display of numerical simulation result on WebGIS. Preliminary results showed that such an integrated approach of combining WebGIS, database and numerical modeling provided effective solutions as needed in an oil spill contingency.