Sample identification method and device

Abstract

The invention provides a sample identification method and device, and relates to the field of data mining. According to the invention, on one hand, the target function is constructed based on the difference relationship between the inter-class divergence matrix and the intra-class divergence matrix, so that the inverse matrix operation of the intra-class divergence matrix is avoided in the projection matrix solving process, and when the sample feature dimension is greater than the sample number, the optimal solution of the projection matrix can still be solved; and on the other hand, through aweighted adaptive mechanism of the inter-class divergence matrix and the intra-class divergence matrix, the optimal weighting coefficients of the inter-class divergence matrix and the intra-class divergence matrix are automatically searched, so that the overall covariance matrix is estimated more accurately, and the projection matrix obtained through solving is more accurate.

Description

technical field [0001] The present disclosure relates to the field of data mining, in particular to a sample identification method and device. Background technique [0002] With the development of information technology, high-dimensional data such as genetic data and image data are generated. [0003] Feature extraction method is one of the methods for mining high-dimensional data. The feature extraction method projects the original high-dimensional features to the low-dimensional space through the projection matrix, thereby reducing the dimensionality of the original features. [0004] Linear Discriminant Analysis (LDA) is a supervised feature extraction method whose goal is to obtain a linear projection matrix W∈R D×d Make based on the between-class scatter matrix S b and within-class scatter matrix S w The value of the objective function tr() constructed by the ratio relationship of is the largest, and the formula is expressed as: [0005] [0006] Among them, max...

AI Technical Summary

Problems solved by technology

[0007] The inventors found that: when the sample feature dimension is greater than the number of samples (i.e. small sample problem), the intra-class scatter matrix S w irreversible, leading to Does not exist, the optimal solution of the linear projection matrix W cannot be obtained

Images