Enabling Deep Learning on Embedded Systems for IoT Sensor Data Analytics: Opportunities and Challenges
Tharmakulasingam Sirojan, B.T. Phung, Eliathamby Ambikairajah · 2018
Internet of Things (IoT) applications are generating enormous amount of data and demands lower latency responses. Due to the wide-spread deployment of IoT sensors, it is impossible to centralize the data processing to produce real-time responses. Gradually, IoT data analytics are pushed out form the cloud servers and performed in decentralized infrastructures. Usage of embedded hardware platforms are constantly increased to facilitate distributed IoT data analytics. Execution of computation intensive data analytic techniques such as deep learning algorithms need to be optimized based on the constraints of embedded hardware platforms. This paper outlines the opportunities and challenges in enabling deep learning on embedded hardware and summarizes the optimization techniques for deep learning approaches.