Data screening method, device, equipment and readable storage medium

Abstract

The invention discloses a data screening method, comprising the following steps: the first participant of the present invention receives the data detection model sent by the coordinator, and detects the data set owned by the first participant based on the data detection model, A reconstruction error set corresponding to the data set is obtained, and then based on the reconstruction error set and the error range configured by the coordinating party, a data screening result corresponding to the first participant is obtained. The invention also discloses a device, equipment and a readable storage medium. Use the data detection model to detect the data sets owned by the participants, so as to screen out the participants and training data that have the same statistical distribution as the training set of the data detection model for federated training. The training data of these participants are similar but different. It can make full use of the diversity of training data owned by the participants, maximize the advantages of federated learning, and train better models.

Description

technical field [0001] The present invention relates to the technical field of machine learning, in particular to a data screening method, device, equipment and readable storage medium. Background technique [0002] In horizontal federated learning, if the training data owned by other participants cannot help themselves, then the participant will not gain anything from participating in horizontal federated learning. Therefore, before training the horizontal federated learning model, you need to choose Participants in horizontal federated learning. [0003] In the prior art, there is a common scheme: the coordinator randomly selects the participants of the horizontal federated learning from the candidate participants (or potential participants), expecting to select training data with relatively balanced statistical distribution. This scheme is very simple, but there is no guarantee that training data with a relatively balanced statistical distribution can be selected. Becau...

AI Technical Summary

Problems solved by technology

[0005] The main purpose of the present invention is to provide a data screening method, device, equipment and readable storage medium, aiming at solving the problem that in the existing horizontal federated learning, the coordinator cannot select good training data or participants' technology for horizontal federated learning question

Images