Classification method and system based on federated few-sample network model and electronic equipment

Abstract

The invention discloses a classification method and system based on a federated few-sample network model and electronic equipment. The method comprises the following steps that: a server acquires an image to be classified, and initiates a judgment request to each client; each client judges the state parameters of the client according to the judgment request and then feeds back a response signal whether the client can participate in the classification task to the server; the server distributes the to-be-classified images to target clients capable of participating in the classification task according to the feedback response signals; each target client inputs the to-be-classified image into a pre-trained few-sample network model for classification to obtain a first classification result; andthe server summarizes and organizes the first classification result and outputs a second classification result. According to the method, the models of the clients only needing a small amount of labeldata are utilized, the problems that in existing machine learning, data privacy is likely to be maliciously attacked and polluted, and a large amount of label data is needed are solved, and good classification accuracy and classification confidence are achieved.

Description

technical field [0001] The invention belongs to the technical field of machine learning, and in particular relates to a classification method, system and electronic equipment based on a federated few-sample network model. Background technique [0002] Artificial intelligence has developed very rapidly in recent years, but the lack of labeled data and the threat of data privacy are still two challenges facing the field of artificial intelligence. On the one hand, due to the value and sensitivity of data, most of the data in the row still exists in the form of isolated islands for the sake of company profits or protection of user privacy, and it is difficult to share the data; on the other hand, the data required by machine learning It is difficult to obtain labeled data, and the lack of labeled data or very few labeled data is common; in addition, the attacker will deduce the input data from some output data of a given model, and may even restore the original training data se...

AI Technical Summary

Problems solved by technology

[0002] Artificial intelligence has developed very rapidly in recent years, but the lack of labeled data and the threat of data privacy are still two challenges facing the field of artificial intelligence

On the one hand, due to the value and sensitivity of data, most of the data in the row still exists in the form of isolated islands for the sake of company profits or protection of user privacy, and it is difficult to share the data; on the other hand, the data required by machine learning It is difficult to obtain labeled data, and the lack of labeled data or very few labeled data is common; in addition, the attacker will deduce the input data from some output data of a given model, and may even restore the original training data set, thereby stealing data, resulting in privacy data leakage

Images