Zero trial learning method and device based on semantic knowledge graph propagation

The invention belongs to the technical field of machine learning, and discloses a zero trial learning method and device based on semantic knowledge graph propagation. Visible sample data and unseen sample data are acquired in an ImageNet data set, a model is trained based on the CNN model, a modification cost function and an aggregation loss function, and the trained CNN model is obtained; setting an optimization function based on the GCN model, adding the feature constraint item to the optimization function to obtain a CGCN model, and training the CGCN model by adopting a self-supervision mode to obtain a trained CGCN model; and setting a final loss function based on the AE model, the matching loss function and the constrained loss function, and training the AE model based on the final loss function to obtain a trained AE model. The problem of distribution drift in the zero trial learning process is relieved, and the accuracy of zero trial learning model verification is improved. 本申请属于机器学习技术领域，公开了一种基于语义知识图传播