Method and device for constructing Gaussian process multi-classifier

Abstract

Embodiments of the invention disclose a method and a device for constructing a Gaussian process multi-classifier and relate to the technical field of electronic information. Through adoption of the method and device, time for classification can be reduced, and training balance is improved. The method comprises the steps of combining every two training samples, and training each combination through a Gaussian process a binary classification algorithm so as to obtain a Gaussian binary classifier, wherein each combination comprises two training samples and is corresponding to a Gaussian binary classifier; performing classification determination on test samples by each Gaussian binary classifier obtained by running; and extracting a category result of test samples from all determined categories, wherein the number of the Gaussian binary classifiers determining the category result of the text samples is greater than the number of the Gaussian binary classifiers determining other categories. The method is applicable to training and test of the Gaussian classifier.

Description

technical field [0001] The invention relates to the technical field of electronic information, in particular to a construction method and device of a Gaussian process multi-classifier. Background technique [0002] In topics such as machine learning and pattern recognition, classification and regression problems are important research directions. The most commonly used classification method is the classification algorithm based on kernel function, especially the Gaussian process classification algorithm based on kernel function. Specifically, the Gaussian process model is a Bayesian machine learning model based on kernel functions and probability discrimination, which can be effectively applied to solve regression and classification problems. Compared with other kernel function classifiers, the advantage of the Gaussian process classifier is that it uses a probability model, and the output is a probability rather than a definite value; and the Gaussian classifier is a param...

AI Technical Summary

Problems solved by technology

[0004] However, the existing one-to-many Gaussian process multi-classifiers need to use all training samples as input when training each binary classifier, so the construction time of the classifier is very long, resulting in a long time-consuming for the entire classification process; in addition, One-to-many Gaussian multi-classifiers will have poor training balance when there are more training sample categories

Images