Training sample selection method based on skew factor model

Abstract

The invention relates to a training sample selection method based on a skew factor model, and the method comprises the steps: carrying out the standardization of an observation data matrix, obtaining a standard matrix, and obtaining a correlation matrix between samples; calculating eigenvalues of the correlation matrix and corresponding eigenvectors thereof, calculating to obtain an accumulated contribution rate, and obtaining the number of main eigenvalues; constructing an eigenvalue matrix and an eigenvector matrix according to the number so as to calculate an initial factor load matrix and normalize the initial factor load matrix; processing the absolute value of each element of the normalized matrix to obtain a first matrix; calculating to obtain a second matrix, and then obtaining a skew reference matrix; normalizing the inverse matrix of the skew reference matrix according to rows to obtain a skew transformation matrix; and finally, calculating to obtain a skew load matrix. According to the method, the collected and labeled training samples are greatly reduced, so that the cost and the period of collecting and labeling the training samples are greatly reduced, and the accuracy of image recognition or classification is further improved while a large amount of manpower and material resource cost is saved.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and in particular relates to a training sample selection method based on an oblique factor model. Background technique [0002] At present, deep learning methods have developed rapidly in the fields of computer vision and artificial intelligence, and have achieved great success in practical applications. However, compared with the traditional "shallow learning method", the deep learning method requires a large number of training samples to train thousands of parameters in the deep learning network to achieve relatively ideal results. The collection and labeling of a large number of training samples will increase huge human or economic costs in practical applications, and the quality and quantity of training samples are directly related to the final effect of learning. Therefore, how to reduce the number of training samples as much as possible, or how to make the required training...

AI Technical Summary

Problems solved by technology

[0005] In view of this, the purpose of the present invention is to overcome the deficiencies of the prior art, to provide a training sample selection method based on the oblique factor model, to solve the problems in the prior art due to the collection and labeling of a large number of training samples. A huge human or economic cost has been added to the problem, and the stability of the classification accuracy cannot be guaranteed

Images