In the training of KGCN, how are the project feature vectors and user feature vectors updated, and how is the loss of the model updated?
I’m having trouble reading the KGCN open source code used for recommendations.In the train method of the source code, the following statements were used for training the model: