为了提高经典 Rosenblatt感知器算法的分类能力 ,该文提出一种基于核函数的非线性感知器算法 ,简称核感知器算法 ,其特点是用简单的迭代过程和核函数来实现非线性分类器的一种设计 .核感知器算法能够处理原始属性空间中线性不可分问题和高维特征空间中线性可分问题 .同时 ,文中详细分析了其算法与径向基函数神经网络、势函数方法和支持向量机等非线性算法的关系 .人工和实际数据的计算结果表明 :与线性感知器算法相比 ,核感知器算法可以有效地提高分类精度 . In order to improve the classification ability of the classic Rosenblatt sensor algorithm, this paper proposes a kernel-based nonlinear sensor algorithm called kernel sensor algorithm, which is characterized by a simple iterative process and kernel function to implement the nonlinear classifier A design kernel sensor algorithm can deal with the linearly separable problem in the original attribute space and the linear separable problem in the high dimensional feature space.At the same time, the paper analyzes the algorithm in detail with the radial basis function neural network, the potential function method and the support vector Machine and other nonlinear algorithms.The artificial and practical data results show that compared with the linear sensor algorithm, the kernel sensor algorithm can effectively improve the classification accuracy.