Incremental learning distributed computing method, system and node based on block chain

Abstract

The invention relates to the technical field of blockchains, and discloses a blockchain-based incremental learning distributed computing method, which comprises the following steps that: a data collection device acquires user data; the server initiates a model training request based on incremental learning, and divides the user data into a plurality of sub-data sets based on a data parallel optimization strategy; the task allocation server divides the model training content into a plurality of model training task calculation layers based on a multi-layer parallel model training method, and publishes the model training task calculation layers to the block chain; each block chain node receives a plurality of model training task calculation layers issued by the task allocation server, and performs model training based on a block chain voting rule; and the block chain combines the model training results stored in each block chain node into a final calculation result according to the divided task calculation hierarchy sequence. The invention further provides an incremental learning distributed computing system and a blockchain node. According to the invention, incremental learning calculation is realized.

Description

technical field [0001] The present invention relates to the technical field of block chains, in particular to a block chain-based incremental learning distributed computing method, system and nodes. Background technique [0002] With the rapid development of science and technology such as the Internet and the Internet of Things, the speed at which humans generate data has increased dramatically. A large number of business decisions and production activities are dependent on data, and it is a hot research topic today to mine the required information from these data through certain means. Machine learning is a key technology to utilize the value of data. Through the study of massive known case data, machine learning can find out the laws that are difficult for humans to discover, and realize the prediction of the future. [0003] In traditional machine learning computing, in order to discover the value of user data, enterprises or institutions need to collect, store, and proc...

AI Technical Summary

Problems solved by technology

[0006] The present invention provides a block chain-based incremental learning distributed computing method, system and node, through the use of distributed computing technology to improve the calculation speed of machine learning content, while using the combination of block chain and incremental learning , which can effectively solve the problems of the decline of model generalization ability in traditional machine learning algorithms and the inability of users to control the transfer of personal data value, while effectively reducing the pressure on centralized servers

Images