A CRITICAL REVIEW ON AI ENABLED IOT SERVICES
Amit Kumar Sharma · Journal of Emerging Technologies and Innovative Research · 2018
In a coming years, billions of linked gadgets will be deployed around our homes, towns, cars and industries in the globe. Devices with limited resources interact with the environment and individuals around them. Many of these gadgets use machine learning models to decipher significance and behaviour underlying the data of sensors, to execute precise predictions and make judgments. The bottleneck is the high amount of linked elements that may congestion the network. Therefore, intelligence on end devices must be incorporated with machine learning algorithms. The deployment of machine learning on such edge devices reduces the network congestion by enabling computation near the data sources. The objective of this work is to evaluate key strategies guaranteeing the execution of machine-learning models on hardware with low performance in the paradigm of the Internet of Things, paving the path to the Internet of Conscious Things. In this work we give a complete assessment of models, architecture and requirements for solutions implementing cutting edge machine learning on the internet of Things devices, with the primary purpose of defining state-of-the-art and imagining development needs. An example of the implementation of edge machine learning on a microcontroller, widely known as Hello World machine learning, is presented.