核矩阵分析法用于支持向量机性能优化  

Analysis of Kernel Matrix Used in Performance Optimization of Support Vector Machines

作　　者：朱树先[1] 朱学莉[1] 郭胜辉[1] 祝勇俊[1] 

机构地区：[1]苏州科技学院机电工程系,江苏苏州215009

出　　处：《控制工程》2012年第6期952-954,共3页Control Engineering of China

基　　金：住房和城乡建设部研究开发项目(2011-K1-19);苏州高新园区培养计划基金项目(20897953)

摘　　要：在学习机的实际应用中,最优核函数的选择是支持向量机的核心问题,直接关系到学习机性能的好坏。为此,通过对核矩阵一些特性的计算和研究,从理论上对常用的核函数性能进行分析和预测。提出了在特征空间中影响核函数性能的一些重要参数,并推导出这些参数的计算方法,为支持向量机核函数寻优提供了一定的理论指导。在此基础上,通过实验仿真证实了利用该方法优选后的核函数确实表现出更加优异的性能。The integrated automation system based on intelligent technologies consisting of three layers： Intelligent Optimization Con- trol, Process Control and Process Management is developed to fit the characteristic and requirement of the Shaft Furnace Roasting process. The structure, functions and control strategy of this system are discussed. Overall production index is achieved using intelligent optimal setting model, fault diagnosis model, intelligent prediction model and loop control. The system proposed in this paper has been successfully applied to the roasting process of a Minerals Processing Factory to realize process optimal control, process optimal operation and process optimal management. It has been proven to provide benefits to this minerals processing factory.

关 键 词：支持向量机 核函数 核矩阵 统计学习理论 结构风险 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]