Federal learning method and system based on graph structure learning, terminal and medium

The invention discloses a federal learning method and system based on graph structure learning, a terminal and a medium, and the method comprises the steps: sampling a plurality of target user sides from all user sides to participate in the training of the current round during the training of each round, updating the local model parameters of the target user sides according to the global model parameters, and carrying out the training of the current round; the target user sides iteratively optimize the local model to obtain optimized local model parameters, then a graph network model is adopted to learn heterogeneity between the target user sides, the optimized local model parameters of all the target user sides are aggregated according to the heterogeneity, global model parameters are updated, and iteration is carried out circularly until optimization of the model is completed. By sampling the user sides to reduce the number of the user sides participating in training, the communication overhead of each roun