【摘要】 支持向量机的关键在于获取分离超平面,一般是基于规划的学习算法,把求解问题转化成凸二次规划问题.采用感知机的迭代算法思路求解支持向量机的分离超平面,算法包括两个迭代过程,首先利用感知机的迭代算法获取一个分离超平面作为初始分离超平面;然后将初始分离超平面不断地旋转和平移,直至正类集和负类集的支持向量到分离超平面的距离相等,此时的分离超平面就是支持向量机的分离超平面.最后分别采用解凸二次规划方法和迭代算法对鸢花分类数据进行学习,获得的支持向量和分离超平面完全一致,说明该迭代算法简单有效.更多还原

【Abstract】 Support vector machine(SVM) is a planned learning algorithm,the key point of which is to obtain the separating hyperplane,and turn the question into convex quadratic programming.This paper solves the separating hyperplane of support vector machine by perceptron iterative algorithm.The solution is divided into two phases.First,the iterative algorithm of perceptron is utilized to obtain an initial separating hyperplane.Second,the initial hyperplane should be rotated and moved continuously until the distance between positive support vector and the separating hyperplane is equal to that between negative support vector and the separating hyperplane.Then the separating hyperplane of SVM can be obtained.The end of paper explains a test that is made on Iris L.,classified data by using the iterative algorithm and convex quadratic programming.The result is completely consistent with that of the support vectors.Therefore,it is proven that the iterative algorithm in this paper is not only concise but also effective.更多还原