Bayesian Deep Learning Models for Recommendation Applications
by Dit-Yan Yeung
While deep learning models have achieved unprecedented success in some challenging perception tasks, probabilistic graphical models are still preferred for many other tasks that require more complex inference mechanisms. To get the best of both worlds, we have been exploring a research direction called Bayesian deep learning (BDL) which seeks to tightly integrate deep learning and probabilistic graphical models within a principled probabilistic framework. In this talk, some BDL models which we have proposed recently for recommendation applications will be presented. Some follow-up work and outstanding research issues for BDL models will also be discussed.
About the speaker: Dit-Yan Yeung is a Professor in the Department of Computer Science and Engineering of the Hong Kong University of Science and Technology (HKUST), with joint appointment in the Department of Electronic and Computer Engineering. His main research interests have been in computational and statistical approaches to machine learning and artificial intelligence, beginning with his doctoral thesis research in neural networks and robotics when he was in the University of Southern California (USC). On the development of machine learning models and algorithms, he has done research on different machine learning approaches including kernel methods, probabilistic graphical models, and neural network models as well as their theoretical connections. Apart from pursuing theoretical research, he has also been interested in developing machine learning models for various applications particularly in computer vision, recommender systems, and, more recently, learning analytics. He publishes frequently in top conferences in machine learning, artificial intelligence, and computer vision.