Heterogeneous federated learning method based on spatiotemporal data distillation

Abstract

The application discloses a heterogeneous federated learning method based on space-time data distillation, which is used for solving the problem that related technologies cannot fully solve the time problem of knowledge reservation, resulting in knowledge forgetting when training a new global model, and the unique space-time characteristics of each client data are not fully considered, so that the efficiency and convergence of federated learning are poor. The server receives the local encoding data obtained by encoding based on the local data sent by each client for data-free global distillation, obtains a first global model, and distributes the first global model to each client; the client performs partial parameter local distillation and non-real-time distillation according to the local data and the first global model combined with dynamic temperature adjustment optimization, generates a local model, and sends the local model back to the server; the server updates the first global model according to each local model through a data-free global distillation strategy, obtains a second global model, and distributes the second global model to each client for next round training.

Images