Fast Decentralized Data Analytics in IoT Wireless Networks

This paper presents an innovative mechanism for decentralized data analytics in the Internet of Things (IoT) networks. In contrast to traditional centralized (eg. Cloud) based solution, our proposed decentralized framework can avoid the problems raised in data collection such as the constraint in high latency, and the associated data breach and privacy issues. The invented mechanism enables each IoT node in the network to independently obtain the global optimal analytics with local computation and communication. The proposed framework has been tested in seismic imaging and machine learning applications. The evaluation results validate its fast convergence to the optimal model, and demonstrates that it allows (near) real-time data analytics in IoT networks under bandwidth and energy limitation, particularly in uncertain and dynamic environments.