告题目 1：Short  Survey on Research at CS@HKR (Computer Science at Kristianstad  University)

报告题目2:  Edge  Machine Learning for Energy Efficiency of Resource Constrained IoT  Devices

报告时间：2019年6月11日（星期二）下午 1：45-3：00

报告地点：计算机学院三楼会议室

主讲人：Dr. Daniel  Einarson and Dr. Dawit  Mengistu, 瑞典克里斯蒂安斯塔德大学

欢迎广大师生参会交流！

With  advances in wearable devices, several IoT applications have emerged in  healthcare for monitoring patients remotely to get insights on symptoms or  trends, and provide better treatment.  However, energy efficiency is a major  challenge in the adoption of wearable IoT because most devices used in these  applications are energy constrained, often operated with low capacity  batteries.   

The  recent shift in machine learning towards the edge offers a new opportunity to  realize intelligent applications on resource constrained IoT devices.  This work  presents a pre-trained recurrent neural network (RNN) model optimized for an IoT  device based on 8-bit microcontrollers.  The model facilitates smart data  transfer operations to improve the energy consumption of wearable devices  equipped with inertial sensors.  

Application  specific optimizations were applied to deploy and execute the pre-trained model  on a device which has only 8KB static RAM storage for the entire parameter set  of the model as well as the sensor data needed for prediction.  Experiments show  that the resulting edge intelligence can reduce the communication cost and  achieve a large saving in energy use (up to 85%).