Stochastic Cumulative DNN Inference for Intelligent IoT Applications

发布日期:2023-04-12      浏览次数:111

主讲人:庄卫华 教授
讲座时间:2023-04-14 09:30:00


主讲人简介:Weihua  Zhuang is a University Professor and a Tier I Canada Research Chair in  Wireless Communication Networks at University of Waterloo, Canada. Her  research focuses on network architecture, algorithms and protocols, and  service provisioning in future communication systems. She was the  Editor-in-Chief of the IEEE Transactions on Vehicular Technology from  2007 to 2013, General Co-Chair of 2021 IEEE/CIC International Conference  on Communications in China (ICCC), Technical Program Chair/Co-Chair of  2017/2016 IEEE VTC Fall, and Technical Program Symposia Chair of 2011  IEEE Globecom. She is an elected member of the Board of Governors and  the President of the IEEE Vehicular Technology Society. Dr. Zhuang is a  Fellow of the IEEE, Royal Society of Canada, Canadian Academy of  Engineering, and Engineering Institute of Canada.

Artificial  intelligence models will continue to be pervasively deployed to support  diverse intelligent Internet of Things (IoT) applications in the 5G/6G  era. Many such applications rely on deep neural networks (DNN) for  object classification. In this presentation, DNN inference uses a  pre-trained DNN model to process an input data sample such as raw  sensing data, and generates a classification result. We will discuss  when to offload DNN inference computation from resource constrained IoT  devices to the edge and how to incorporate different contributions from  multiple random DNN inference results to improve task classification  accuracy, while achieving high transmission, computation, and energy  resource utilization.