Model training, data classification method, device, electronic device and storage medium

Abstract

The invention relates to a model training, data classification method, device, electronic equipment and storage medium. The method includes: obtaining a D-dimensional sample embedding vector and a sample category of a training sample; using the D-dimensional category embedding vector and the D-dimensional sample embedding vector as the depth The input of the matching model is matched, and the matching score between the training sample and the D-dimensional category embedding vector is output; the intra-class max layer determines a matching score among the M matching scores corresponding to each preset category, which is used as the matching score between the training sample and each preset category. Category matching score; the inter-class softmax layer calculates the probability that the training sample belongs to each preset category based on the category matching score of the training sample and each preset category; using the probability that the training sample belongs to each preset category and the sample category, the depth The classification model is trained until the deep classification model converges, and the deep classification model is obtained. In the embodiment of the present invention, by using M D-dimensional category embedding vectors to describe categories, the classification error is reduced and the classification is more accurate.

Description

technical field [0001] The present invention relates to the field of computer technology, and in particular, to a model training, data classification method, device, electronic device and storage medium. Background technique [0002] The classification problem is a common problem in business. Under the premise of having accurate and sufficient training data, classification models based on deep learning can usually achieve good results and become commonly used models. [0003] When classifying videos according to their titles, if one of the preset categories is a game category, because in practical applications, the content of the game category is rich and complex, and the more related videos, the more large-scale game-related videos occupy the sample data. Most of the proportions, the other parts of the sample data are some video related to small games. If a D-dimensional category embedding vector is used to represent the game category, then this D-dimensional category embe...

AI Technical Summary

Problems solved by technology

[0004] Like this, when A small game related video is classified, because the distance between the D-dimensional sample embedding vector of A small game related video and the mean vector is relatively far, the small game related video is classified into the X category (the X category The vector distance between the mean vector of the D-dimensional category embedding vector and the distance between the D-dimensional sample embedding vector is relatively close), so there will be classification errors, and the classification error rate is high

Images