Functional Bayesian Deep Learning:

Beyond Function Approximation to Function Distribution Approximation

Dr. Junyu Xuan

DECRA Fellow, and Senior Lecturer of AAII at FEIT

Abstract
Bayesian Deep Learning (BDL) is a burgeoning field that merges the potent function approximation abilities of deep learning with the uncertainty modeling strengths of traditional Bayesian methods. This combination promises to improve model generalization and resilience, providing valuable uncertainty assessments for various applications like critical medical diagnostics and diabetes detection. However, this fusion also introduces complexities to traditional posterior inference in parameter space, including ambiguous priors, intricate posteriors, and potential anomalies. This presentation will delve into the motivations, principles, and techniques behind BDL in function space, transitioning to key technological advancements and their roles in machine learning tasks. Finally, we will explore the current challenges encountered by BDL.

Biography
Dr Junyu Xuan is an IEEE/ACM Senior Member, ISBA/BNP Life Member, ARC Discovery Early Career Researcher Award (DECRA) Fellow, and Senior Lecturer of Australia Artificial Intelligence Institute in the Faculty of Engineering and IT at the University of Technology Sydney (UTS). His research interests include Probabilistic Machine Learning, Bayesian Nonparametric Learning, Bayesian Deep Learning, Reinforcement Learning, Text Mining, Graph Neural Networks, etc. He has published over 60 papers in high-quality journals and conferences, including Artificial Intelligence Journal, Machine Learning Journal, IEEE TNNLS, IEEE TKDE, ACM Computing Surveys, ICDM, NeurIPS, AAAI, etc. He served as PC or Senior PC member for conferences, e.g. NIPS, ICML, UAI, ICLR, AABI, IJCAI, AAAI, EMNLP, etc.

More information: https://profiles.uts.edu.au/Ju...