Federal learning optimization method and device

Abstract

The invention provides a federal learning optimization method and device. The method comprises the steps: obtaining a global model and a delayed global gradient which are transmitted by a server side in the current round of federated learning, wherein the delayed global gradient is obtained by updating the global gradient of the last round based on respective local data in the last round of federal learning; on the basis of the global model and the delayed global gradient of the current round, updating the local model through the local data, obtaining the federated learning update quantity, wherein the federal learning update quantity comprises the update quantity of the local model and the update quantity of the delayed global gradient; sending the federal learning update quantity to the server side. Therefore, the server side performs information aggregation according to the federated learning update quantity to obtain a new global model and a new global gradient, and sending the new global model and the new global gradient to each client side for next round of federated learning. The method effectively overcomes the problem of model deviation, improves the communication efficiency, and reduces the calculation complexity.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a federated learning optimization method and device. Background technique [0002] Federated Learning (Federated Learning) refers to the establishment of a predictive model by combining multiple distributed clients when data is distributed across multiple clients (such as edge devices, mobile terminals, and servers) and not shared. The distributed machine learning paradigm has therefore become a new solution with great potential to break the "data island", ensure that the data of each participant is always out of the local area, and effectively aggregate the information of each participant. figure 1 For a schematic diagram of the federated learning update steps provided by the present invention, please refer to figure 1 As shown, federated learning can be obtained based on the improvement and upgrading of traditional data centralized cloud computing platforms, so...

AI Technical Summary

Problems solved by technology

It can be seen that the model deviation in the existing federated learning will affect the convergence

Images