Semi-supervision classification method based on flow shape alignment

Abstract

The invention discloses a semi-supervision classification method based on flow shape alignment. The semi-supervision classification method includes the steps of establishing class coordinates for training data of an auxiliary domain and a target domain based on known class information, obtaining the similarity between the training data in the target domain and test data and the similarity between the test data based on a target domain neighbor graph, calculating the similarity between the training data in the auxiliary domain and the test data in the target domain by discovering the relevance between the data of the auxiliary domain and the target domain, calculating the class coordinates of test data in the target domain based on a semi-supervision model, and obtaining the class information of the test data according to the class coordinates thereof. The method uses data with known classes in the auxiliary domain to classify data in the target domain, thereby improving classification precision in the target domain.

Description

technical field [0001] The invention belongs to the field of pattern recognition, and relates to a machine learning method based on semi-supervised classification, in particular to a semi-supervised classification method based on instance-level manifold alignment. Background technique [0002] Traditional classification algorithms need enough labeled data to train high-precision classifiers, and require the training data and target data to come from the same data domain. When faced with the classification of data on a new data domain, traditional classifiers cannot use the labeled data on the existing data domain to classify the new data. In order to obtain enough training data, it is often necessary to manually analyze and label the data. Doing so is undoubtedly a waste of existing data information, and at the same time requires a lot of manpower, material resources and time, which is not only inefficient, but also expensive. In recent years, the problem of how to use the...

AI Technical Summary

Problems solved by technology

However, there is often no corresponding point information or coordinate information between datasets with only category information, which brings great challenges to manifold alignment

Images