Classifier training method and device based on semi-supervised learning

Abstract

The invention relates to the technical field of machine learning, in particular to a classifier training method and device based on semi-supervised learning. The method includes: when a model training instruction is received, obtaining an initial classification model and each sample data carrying a positive label, sample data carrying a negative label, sample data carrying a third label, and unlabeled sample data, and generating a training Data set, input the training data set into the initial classification model to trigger its training, obtain the target classification model after training, and calculate the classification accuracy; when the classification accuracy reaches the preset accuracy threshold, determine the target classification model as a classifier. Through various labeled sample data and unlabeled sample data, more classification and identification information is provided for the initial classification model. The initial classification model can learn the corresponding label of each data according to each sample data, so as to improve the classification of each data. Recognition and classification accuracy, improving classifier performance.

Description

technical field [0001] The present invention relates to the technical field of machine learning, in particular to a classifier training method and device based on semi-supervised learning. Background technique [0002] In the field of machine learning technology, in order to train a classifier with good classification performance, it needs to be trained through a large amount of sample data. Semi-supervised learning is a machine learning method that combines supervised learning and unsupervised learning. It performs multiple iterative training given a large number of positive label data, negative label data and unlabeled data to obtain a classifier with good performance. Among them, the classifier is mainly used for data classification, that is, according to the binary classification problem, it is determined whether the data to be classified is a positive class or a negative class. Therefore, machine training through semi-supervised learning can improve the classification ...

AI Technical Summary

Problems solved by technology

[0003] When the classifier is used for character recognition, the character recognition of the classifier is still not optimal, for example: the similarity between the letter "O" and the number "0" is extremely high, and the character "O" is recognized by classification When it is easy to recognize the number "0", it is impossible to classify letters and numbers correctly, and if the characters to be classified and recognized by the classifier are handwritten characters by the user, since each user's writing style is different, if the character is written too much If it is scribbled, it cannot be recognized, and characters with many strokes tend to have too many continuous strokes during the writing process, making the recognition success rate of the machine very low when recognizing the user's handwritten characters

Images