关键词: 感知器, 解向量, 法向量, 最小平方误差法, 变步长学习法

Abstract: This paper presents a new algorithm based on the Shortest Distance of perception，and puts forward the process by analyzing the relation between soluble vector and normal vector.Simultaneously，it compares the new algorithm proved by examples with the LMSE and the learning algorithm of variable paces.Consequently，on linearity separable，it shows that the new algorithm SDA can solve convergence in all the traditions and enhance the speed of calculation.

Key words: the perception approach, soluble vector, normal vector, the limited square error, the learning algorithm of perception with variable paces