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