Generalization Performance Of Fisher Linear Discriminant Based On α-mixing Sequence

Fisher linear discriminant method is known as one of the methods of dimensionality of classification, Many scholars are worth to Fisher linear discriminant method of theory and application research. While for Fisher linear discriminant method up to now, generalization performance research is mainly based on the sample points is independent identically distributed assumptions. And assuming no matter in theory (independent identically distributed, or in the practical application is very strong. In particular, in the practical application of machine learning, Many data are independent identically distributed, Fisher linear discriminant method based on independent identically distributed application of theory to study the learning performance of independent identically distributed data, such as convergence rate is inappropriate.In this article, we will based on Fisher linear discriminant method independent identically distributed theory is extended to the independent identically distributed data, that is, data is a sequence of mixing. Our main research based on a mixing sequences of Fisher linear discriminant method of generalization performance and convergence rate. And were compared with known results. In this paper, the main content is as follows:First introduced the research background and significance of Fisher linear discriminant method Also this paper introduces the definition and related properties of mixed sequence. Then introduce independent identically distributed situation Fisher linear discriminant method of generalization performance. Finally, we established based on a mixed Sequence of the Fisher linear discriminant method of generalization performance, convergence rate, such as theory, Prove based on a mixed data sequences Fisher linear discriminant method is the same. And the results are generalized to other sequence of mixed.