Data processing method and device, equipment and medium

The invention discloses a data processing method and device, equipment and a medium. The method comprises the following steps: acquiring a plurality of training samples, respectively inputting the current training samples into to-be-trained models, and determining gradient parameters of the to-be-trained models; obtaining each gradient parameter belonging to the same participant to obtain a first fusion gradient parameter; determining a to-be-updated gradient parameter based on the first gradient fusion parameter corresponding to each participant, and updating the parameter of each to-be-trained model based on the to-be-updated gradient parameter, so as to train each updated to-be-trained model based on a next to-be-trained sample when the next to-be-trained sample is received; and when it is detected that the loss function in the to-be-trained model converges, training to obtain a target model deployed to each participant, so as to process the to-be-processed data based on the target model and obtain a targe