Prof.Zhanyu Ma

Speech Title:

Probabilistic Model-based Deep Neural Networks Optimization

Abstract:

The rapid development of artificial intelligence technology represented by deep learning has received a great deal of attention from academia and industry. As an important component of deep learning, deep neural networks suffer from overly complex structures, unclear functional mechanisms such as attention, and incomplete observation data, and the research of their optimization methods is facing challenges. The team focuses on the research of deep neural network optimization methods based on probabilistic model representation, and proposes a deep neural network regularization framework based on non-Gaussian prior, a deep neural network attention mechanism based on non-Gaussian prior, and a deep neural network output mechanism based on hybrid model. We proposed deep neural network regularization framework based on non-Gaussian prior, deep neural network attention mechanism based on non-Gaussian prior, and output feature uncertainty estimation method based on hybrid model. These methods effectively reduce the complexity of the network, better explain the attention mechanism of the model, and improve the confidence level of the prediction results.

Bio of lecturer:

Zhanyu Ma is currently a professor at Pattern Recognition and Intelligent System (PRIS) Lab, School of Artificial Intelligence, Beijing University of Posts and Telecommunications (BUPT), Beijing, China. He received his PhD degree from KTH-Royal Institute of Technology, Stockholm, Sweden, in 2011. From 2012 to 2013, he has been a postdoctoral researcher at KTH. His research interests include pattern recognition and machine learning fundamentals with the applications in non-Gaussian statistical models, computer vision, big data modeling and analysis, multimedia signal processing. He has published over 100 papers in refereed international journals and conference proceedings, including TPAMI, TNNLS, TASLP, PR, CVPR, ICASSP, etc. He serves as an associate editor of IEEE Trans. on Vehicular Technology, IEEE Trans. on Neural Networks and learning Systems, NEUROCOMPUTING (Elsevier), technical co-chair of SPLINE 2016, program co-chair of IEEE MLSP 2018. He is a senior member of IEEE.

Back to Invited Speakers/Panelists home page ↵