Federated model training, customer profiling method, device, equipment and medium

Abstract

The present invention relates to the technical field of user portraits, and provides a federated model training, customer portrait method, device, equipment, and medium. The method includes: obtaining a list of participants and an initial customer portrait federation model, and according to a preset screening plan, from the participants Select qualified participants from the list; send the initial customer portrait federation model to each qualified participant; receive the returned model parameters; use the MPI parallel method to extract abnormal features through the malicious parameter detection model, and output each model parameter according to the extracted abnormal features The identification results of the malicious parameters are filtered to obtain the final normal parameters; updates and federated learning are performed to obtain the global customer portrait federated model. The invention realizes the use of the MPI parallel method to extract abnormal features through the malicious parameter detection model, and filter the malicious parameters, automatically remove the malicious parameters provided by the malicious participants, and improve the efficiency and accuracy of federated learning modeling.

Description

technical field [0001] The present invention relates to the technical field of user portraits, in particular to a federated model training, customer portrait method, device, computer equipment and storage medium. Background technique [0002] With the increasing popularity of machine learning, big data-driven intelligent applications will soon be applied to all aspects of our daily lives, including intelligent voice, medical care, transportation and more. However, in traditional machine learning methods, the key to ensuring the accuracy of the training model is to collect a sufficient amount of data, which may contain private information about individuals, such as personal medical information or personal itinerary information. Public concerns about the compromise of personal privacy. Recently, federated learning has been more and more widely used because of its significant advantages in privacy protection. By directly aggregating the model parameters of each participant, th...

AI Technical Summary

Problems solved by technology

However, in traditional machine learning methods, the key to ensuring the accuracy of the training model is to collect a sufficient amount of data, which may contain private information about individuals, such as personal medical information or personal itinerary information. Public concerns about compromised privacy

Recently, federated learning has been more and more widely used because of its significant advantages in privacy protection. By directly aggregating the model parameters of each participant, the model parameters obtained by the aggregation are used to train the global federated model, and then the training The final global model parameters are then fed back to each participant for each participant to update the local model. However, if there is a malicious participant in each participant and the malicious participant provides a false or malicious model during the local training process Parameters will lead to lower effectiveness of global model parameters, which will directly affect the overall federated model quality, leading to the failure of the entire federated learning process, which will lead to lower efficiency and accuracy of federated learning modeling

Images