Target recognition federal deep learning method based on trusted network

Abstract

The invention discloses a target recognition federated deep learning method based on a trusted network. The method is mainly composed of a local model and a federated model. The local model and the federated model have the same structure, the local model and the federated model are trained by adopting the same optimization algorithm (such as an Adam optimizer) and passing training parameters (such as a learning rate eta, a neural network weight w, a loss function E and the like), the local model and the federated model jointly train a convolutional neural network by adopting a federated learning mode, training data of each client is local, and the data is immobile and the model is mobile. The highest recognition precision of the method can reach 91%, and the method has the advantages of being high in recognition precision and high in convergence speed. Through the method, the problems of difficulty in data fusion, long decision response time and the like in cross-client fields can be solved, the decision time is shortened, and finally, a quick response effect can be realized.

Description

technical field [0001] The invention relates to the technical field of mobile communication, in particular to a trusted network-based federated deep learning method for object recognition. Background technique [0002] Object recognition is the process by which a particular object or type of object is distinguished from other objects or types of objects. It includes both the identification of two very similar objects and the identification of one type of object from another. Object recognition is widely used in various fields of production and life. High-precision object recognition algorithms are usually based on learning and training for a large amount of data. However, in the current legal environment, it is becoming more and more difficult to collect and share data between different organizations, especially those highly sensitive data (financial transactions, medical and health data, etc.), in consideration of privacy and data security, It is easy for data owners to ...

AI Technical Summary

Problems solved by technology

At present, due to the limited amount of data in various fields, it is not enough to support deep learning models for high-precision training; at the same time, the current management systems in various fields are different, and the information systems developed are different, making it difficult to share and exchange data sources across domains. In a sense, the data of each system has become more and more "data islands"

To fully unify the data standards among the various systems, and carry out unified data fusion processing and application, the workload will be huge

Images