Method, device and medium for solving heterogeneous federated learning by using hypernetwork

The invention discloses a method and equipment for solving heterogeneous federated learning by using a super network and a medium, so as to realize the purpose of generating a model weight for a heterogeneous model on edge equipment through the super network and help different convolutional neural network models on equipment such as a camera and a smart phone to realize heterogeneous federated learning under the condition of different model structures. The federated learning task of the model can still be accurately and quickly completed, and the memory space required to be occupied in the federated learning training process is reduced. Comprising the following steps: reasonably dividing a cifar10 data set; generating a heterogeneous model; heterogeneous federated learning is carried out; and S3, the step S2 and the step S3 are repeated, when the client heterogeneous model converges, the super network generates a model weight for the client, the client model can accurately complete a target identification tas