Graph network classification model training method, apparatus and system, and electronic device

Abstract

The invention belongs to the field of graph data security, and particularly relates to a graph network classification model training method, device and system and electronic equipment. The method comprises the following steps: acquiring a graph data set, and dividing the graph data set into label nodes and label-free nodes; inputting the graph data with the label nodes into a classification model for training; using the trained classification model to classify graph data without label nodes; randomly selecting adversarial nodes from label nodes and label-free nodes, and calculating the influence of the adversarial nodes on a loss function of an adversarial sample generator so as to generate an adversarial sample; inputting the adversarial sample into the trained classification model for training again to obtain an enhanced classification model. According to the invention, the trained classification model is used to predict the labels of the label-free nodes, so that the connectivity problem and the low label rate problem in semi-supervised learning can be solved. According to the method, the robustness of the node classification model can be effectively improved.

Description

technical field [0001] The invention belongs to the security field of the graph field, and in particular relates to a training method, device, system and electronic equipment of a graph network classification model. Background technique [0002] In recent years, due to the ubiquity of graph data in the real world, researchers have begun to think about how to apply deep learning models to graph data. [0003] Graph deep learning models are widely used in social networks, community detection, and recommendation systems. Among them, the graph convolutional neural network is the most important branch of the graph deep learning model. The application scenarios of the graph convolutional neural network are generally divided into two categories, one is the task at the node level, and the other is the task at the graph level. Among the tasks at the graph level, the most common tasks are graph generation and graph classification; among the tasks at the node level, the most common t...

AI Technical Summary

Problems solved by technology

At the same time, the present invention can effectively solve the problem of lack of connectivity between the training set and the adversarial samples in the adversarial training

Images