High-quality federal learning system and learning method based on block chain and reputation mechanism

Abstract

The invention belongs to the technical field of federal learning, and discloses a high-quality federal learning system and learning method based on a block chain and a reputation mechanism, and a block chain layer shares a global model and reputation evaluation by taking an intelligent contract as a carrier; the reputation layer scores the local models shared by all the data owners by using a model scoring mechanism; reputation evaluation is carried out on each data owner by using a reputation mechanism; a model training layer data owner carries out training of a local model and shares the local model to a model aggregator; and meanwhile, the model aggregator aggregates the collected models in combination with the reputation and awards participating data owners. The federated learning system based on the block chain disclosed by the invention has important guidance and technical significance for realizing a high-quality model aggregation task, is beneficial to improving the security and robustness of federated learning, and is beneficial to promoting the fairness of reward distribution and the enthusiasm of participating in the task.

Description

technical field [0001] The invention belongs to the technical field of federated learning, and in particular relates to a high-quality federated learning system and learning method based on blockchain and reputation mechanism. Background technique [0002] Currently, federated learning is a machine learning paradigm proposed by Google to solve privacy and security issues in machine learning. Federated learning no longer uses centrally stored data to train models, but requires participants to use their own data to perform model training locally and send model-related parameters or gradients to the central aggregation server. The Model Central Server uses these local models to generate an accurate global model. During this whole process, since the personal data of the participants does not need to be directly involved in sharing, the security of personal data is well protected while obtaining the ideal model. [0003] Federated learning faces a key problem: how to encourage ...

AI Technical Summary

Problems solved by technology

[0005] However, the existing technology does not have a method of building a federated learning system based on blockchain technology, and there is no federated learning system based on blockchain

[0006] Through the above analysis, the existing problems and defects of the existing technology are: the existing technology does not have a method of building a federated learning system based on blockchain technology, nor does it have a federated learning system based on blockchain; at the same time, the existing federated learning system The data security is not high, the quality is not high, and the risk of a single point of failure is high

[0007] The difficulty in solving the above problems and defects is: the existing blockchain technology cannot be directly applied to the construction of the federated learning system; the nodes of the federated learning are difficult to control and predict, and the data quality is difficult to measure

Images