Credible Internet of Intelligence Toward Future Multidimensional Connection

【Author】 Qiu, Chao; Qi, Haotian; Liu, Zhicheng; Wang, Xiaofei; Jiang, Jing; Hu, Qinghua

【Source】IEEE NETWORK

【影响因子】10.294

【Abstract】In the multidimensional future connection, users' demands for high data rate, ubiquitous accessibility, and low latency have fueled the continuous advancement of artificial intelligence applications, especially edge intelligence. Meanwhile, the methodology of edge intelligence has been widely researched. However, there are some challenges remaining to be resolved, including the divisible feature of deep networks and the credibility of edge intelligence nodes. In this article, we consider a credible Internet of Intelligence (IoI) framework, with the help of deep networks segmentation and aggregation, as well as blockchain. Specifically, deep networks are separated into several sub-models to be trained by numerous edge trainers. Meanwhile, smart contract in the blockchain records and calculates the credibility of each edge trainer. According to the credibility, the trained sub-models would be adopted or abandoned. In order to show the effectiveness of the IoI system, we employ it to solve the problem of future multidimensional connection. Experimental results reveal the superior performance of the proposed scheme.

【Keywords】Training; Artificial intelligence; Blockchains; Task analysis; Internet; Training data; Data models

【发表时间】

【收录时间】2022-06-06

【文献类型】实证性文章

【主题类别】

区块链技术-协同技术-边缘计算