Data amplification method based on high-dimensional space transformation and machine recognition system

Abstract

The invention belongs to the technical field of image processing and machine learning, and discloses a data amplification method based on high-dimensional space transformation and a machine recognition system. The method includes the steps of converting the background sample data from the original space to the high-dimensional space, obtaining a high-dimensional space target sample distribution based on a distribution histogram of the background sample, generating high-dimensional space target sample data, using a distance function to conduct equation set transformation, transforming the amplification data from the high-dimensional space to the original space. According to the method, the distribution histogram of the negative samples is learned, and a corresponding positive sample data set is expanded, and the problem that positive and negative sample data in a machine learning model cannot be matched is solved. The classification performance is improved, and the classification accuracy of the positive samples is especially improved. Statistical analysis is carried out on the basis of the background samples to obtain distribution of the target sample data to be generated, so thata target sample is generated, and the effectiveness of the amplification data is improved. The problem of sample overlapping and model overfitting in conventional method when new target samples are synthesized on the basis of a small number of samples can be solved.

Description

technical field [0001] The invention belongs to the technical fields of image processing and machine learning, and in particular relates to a data augmentation method and a machine recognition system based on high-dimensional space transformation. Background technique [0002] Machine learning is a study of machine recognition of existing knowledge and acquisition of new knowledge and new skills. It has been widely used in various fields, such as image recognition, data mining, fault diagnosis, etc. In machine learning technology, the sample data needs to be processed and trained first. In practical applications, the sample data set is often unbalanced. Usually, the number of negative samples in the data set is far more than the positive samples. The result of training on such data sets is that the classification performance of the classifier is reduced; for example, in the problem of vascular plaque recognition In the vascular system samples, vascular plaques tend to accou...

AI Technical Summary

Problems solved by technology

[0003] To sum up, the problems existing in the existing technology are: synthesizing new samples based on the analysis of the target samples, it is easy to cause problems such as sample overlap, ignoring boundaries and isolated points, etc. Due to the limitations of training samples, the classification of the classifier is inaccurate , there are certain limitations in improving the classification performance of target samples, such as sample overlap may cause model overfitting problems, ignoring boundaries and isolated points will cause misclassification of such sample points, etc.

Images