Prof. Tang Niansheng of Yunnan University Delivered Academic Report at BISTU

By Mon, 11 Mar 2024 GMT


Beijing, 11th March 2024 (BISTU)--- Upon invitation, Prof. Tang Niansheng, Second-tier professor, Ph.D. supervisor, and Dean of School of Mathematics and Statistics of Yunnan University, delivered an academic report at BISTU on 8th March 2024.

 

Secretary Xie Jiangling of CPC BISTU Committee met with Prof. Tang Niansheng and extended warm welcome and sincere thanks to his visit and academic lecture.

 

“Deep expectation-maximization network for unsupervised image segmentation and clustering”, Prof. Tang Niansheng’s report, was delivered at Shahe Campus.

 

Prof. Tang offered a detailed explanation of Deep expectation-maximization network.  Deep expectation-maximization network (DEM) is based on Gaussian Mixture Models (GMM) for statistical modeling of images in their latent feature space and is implemented in a new deep learning framework. In the unsupervised setting, an autoencoder network and an EM module are designed on the image potential features for jointly learning the image potential features and the GMM model of the potential features in a single framework. To construct the EM module, the EM algorithm and the online EM algorithm are unfolded in a fixed-step iterative operation into microscopic network blocks that are inserted into the network to estimate the GMM parameters of the image potential features. The proposed network parameters can be optimized end-to-end using losses based on GMM log-likelihood, probabilistic entropy of Gaussian component assignment and image reconstruction error. It is confirmed by extensive experiments that the proposed DEM network achieves good results compared to several state-of-the-art methods in unsupervised image segmentation and clustering.

 

During the meeting, the young teachers of School of Applied Sciences thought actively and spoke enthusiastically, exchanging ideas with Prof Tang Niansheng and discussing the future development in this field. Prof Tang Niansheng answered the questions of the young teachers in an insightful way, and at the same time, he expanded the answers to other relating fields. which truly benefited all participants.

The report was chaired by Vice President Xiao Zhisong and attended by faculty and students of School of Applied Sciences.

About Prof. Tang Niansheng

TANG Niansheng, Ph.D., is the winner of the national science fund for distinguished young scholars, the Changjiang scholar of the Ministry of education, Second-tier professor and Ph.D. supervisor and Dean of School of Mathematics and Statistics of Yunnan University, Vice President of Chinese Association for Applied Statistics, New Century Excellent Talent of Ministry of Education, young and middle-aged outstanding expert of National Hundred, Thousand, and Ten Thousand Talents Project, leading talents of science and technology in Yunnan province, the first batch of Yunling scholars in Yunnan province, academic and technical leaders of young and middle-aged in Yunnan province, an Elected Member of International Statistical Institute, Board of Directors of ICSA , and chief scientist of state key projects. He is associate editor-in-chief and editor of a number of globally renowned journals of statistics such as CIMS. His research interests include Bayesian statistics, missing data analysis, variable selection, high-dimensional data analysis, data science, and machine learning and have published more than 190 papers in Annals of Statistics, JRSSB, JOE, JASA, Biometrika, etc.