基于大规模数据的支撑矢量机的训练和分类

Training and classification of SVMs based on large-scale data

支撑矢量机是一种基于统计学习理论的、新颖且有强的泛化性能的学习方法 ,可看作是一种训练多项式神经网络或径向基函数分类器的新方法 .支撑矢量机可用于模式识别、回归估计、求解线性算子方程等 .介绍了支撑矢量机的分类机理 ,并针对大规模数据讨论其训练和分类中存在的问题及典型的解决方法 .

Support vector machines (SVMs) are a novel learning technique based on statistical learning theory. They have highly generalized ability and can be seen as a new method for training polynomial neural networks or radial basis function classifiers. SVMs can be used in pattern recognition, regression estimation, solving linear operator equation, etc. The classification principle of SVMs is described. The problems which exist when large scale data are trained and classified are presented, and typical solutions are discussed.