关键词: 核最小均方误差（KMSE）, 再生核, 改进的核最小均方差, 鉴别向量

Abstract: According to the reproducing kernel theory，the discriminant vector in the feature space associated with Kernel Minimum Squared Error （KMSE） model can be approximately expressed in terms of a linear combination of samples selected from all of the training samples.This implies that the general KMSE can be improved for more efficient implementation.The criterion is proposed for the first time in the paper，and then an improved kernel minimum squared error algorithm has been developed based on the criterion，and the experiments show that the proposed method not only is simple，efficient，but also has good performance.

Key words: Kernel Minimum Squared Error（KMSE）, reproducing kernel, Improved Kernel Minimum Squared Error, discriminant vector