Data Flow and Distributed Deep Neural Network based low latency IoT-Edge computation model for big data environment

作者：Veeramanikandan; Suresh Sankaranarayanan; Joel J.P.C. Rodrigues; Vijayan Sugumaran; Sergei Kozlov

来源：Engineering Applications of Artificial Intelligence, 2020, 94.

DOI：10.1016/j.engappai.2020.103785

• Big data
• Data in motion
• Data flow processing
• Deep Neural Networking
• Edge computing
• Fog computing
• Internet of Things
• Latency

The trillion-fold increase in computing power brings the accessibility of deep learning to everyone. Deep learning offers precise information almost all the time when compared to other learning algorithms. On the other hand, the popularity of Internet of Things (IoT) has increased in various areas such as Smart City, Oil Mining, and Transportation. Edge/Fog computing environment helps to handle significant challenges faced by the IoT, viz. latency, bandwidth consumption, and everlasting network connectivity. For analytics in Edge computing, which is distributed in nature, the trend is more towards distributed machine learning. This research work is focused on the integration of data flow and distributed deep learning in the IoT-Edge environment to bring down the latency and increase accuracy starting from the data generation phase. To this end, a novel Data Flow and Distributed Deep Neural Network (DF-DDNN) based IoT-Edge model for big data environment has been proposed. Our proposed method has resulted in latency reduction of up to 33% when compared to the existing traditional IoT-Cloud model.