Prof. Tao Gu

Title: Pushing Deep Learning on Mobile Devices



Mobile Deep Learning has emerged as a privacy-preserving learning paradigm for mobile and embedded devices. This paradigm offers unique features such as privacy preservation, continual learning and low latency inference. However, squeezing Deep Learning to mobile devices is extremely challenging due to resource constraint. Traditional Deep Neural Networks are usually over-parametered, hence incurring huge resource overhead for on-device learning. In this talk, I will present two solutions. In the first solution, we leverage on multiple devices and design a decentralized mobile deep learning framework to enable on-device collaborative learning. To address resource challenges, we create several innovations including a chain-directed Synchronous Stochastic Gradient Descent algorithm to effectively reduce resource overhead among multiple devices. In the other solution, we focus on a single device learning framework where we use the Model Predictive Control theory to grow a Deep Neural Network from tiny to backbone to achieve resource efficiency. Both systems are built on Android OS and have been extensively evaluated with a range of datasets. I will also discuss our on-going research and future direction in this field. At the end of my talk, I will present some of my recent work in activity and gesture recognition.


Prof. Tao Gu is currently a Professor in School of Computing at Macquarie University, Sydney. He obtained his Ph.D. in Computer Science from National University of Singapore, M.Sc. in Electrical and Electronic Engineering from Nanyang Technological University, and B.Eng. in Automatic Control from Huazhong University of Science and Technology. His current research interests include Internet of Things, Ubiquitous Computing, Mobile Computing, Embedded AI, Wireless Sensor Networks, and Big Data Analytics. He is currently serving as an Editor of Proc. of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), an Associate Editor of IEEE Transactions on Mobile Computing (TMC) and IEEE Internet of Things Journal (IoT-J). The long-term goal of his research aims to discover innovative ways of sensing and connecting the physical world and embedding AI intelligence to facilitate the building of new applications. Please visit for more information.