The Internet Of Things ( Iot )

1292 WordsDec 13, 20166 Pages
Rochester Institute of Technology Real-Time and Embedded Systems SWEN 563 / CMPE 663 / EEEE 663 Graduate Paper Review Report ON “Efficient Embedded Learning for IoT Devices” Submitted by: Rohan N. Dhamdhere, rnd7528@g.rit.edu I. Introduction The Internet of Things (IoT) helps us sense meaningful data from the physical world. In recent years, its ability to sense the data has increased tremendously, which increases the prospects of IoT applications useful to humans. Machine learning will play a big part to help IoT derive sense from raw data and deliver useful information. A simple quantitative analysis is done on two broad categories of applications – a smart glass that uses a deep learning network to perform real time recognition on video stream that it captures, and a server cluster that trains the model associated with the deep network. The analysis showed differences between the computational capabilities of current systems and the needs levied by these application cases. The authors hypothesize that deep learning will be one of the important part of the computing platforms and will be extensively deployed. The Paper discusses the following potential procedures to improve the efficiency of deep learning for embedded computing systems – accelerators for deep learning, approximate computing and neuromorphic circuits based on emerging post-CMOS devices. II. Computational Challenges The
Open Document