SPATIAL GRAPH CONVOLUTIONAL NETWORK TRAINING METHOD, ELECTRONIC DEVICE AND STORAGE MEDIUM

摘要

A spatial graph convolutional network training method, comprising: acquiring training data (S110), wherein the training data comprises network structure features of multiple objects, the object attribute features of each object, and the mark categories of some objects among the multiple objects, the network structure features of each object are correlations between said object and other objects, and among the multiple objects, objects having mark categories are second objects, and objects not having mark categories are first objects; and training, according to the training data, a graph convolutional network to be trained, so as to obtain a graph convolutional network for object classification and object network structure attribute prediction (S120). The present invention achieves simultaneous processing of tasks of object classification of an object relationship network and network structure feature prediction of the object, thereby saving the computing power of a computing device and improving the efficiency.