Biography:  Fang Fang (Chinese: 方方) is Professor of statistics at East China Normal University and deputy director of the Key Laboratory of Advanced Theory and Application of Statistics and Data Science - MOE. He received his B.S. in Mathematics from Peking Unversity and Ph.D. in Statistics from the University of Wisconsin - Madison. Before joining East China Normal University in 2013, he worked in GE Capital and Shanghai Pudong Development Bank.  He has published nearly 30 papers in world-class statistical journals including AOS / JOE / Biometrika，mainly focusing on missing data, model averaging and deep learning. He is the Associate Editor of Journal of Nonparametric Statistics. 

Title: Generative adversarial nets for fragmentary data imputation and prediction

Abstract: Fragmentary data has become more and more popular in recent years and brings big challenge to traditional missing data methods due to huge missing proportions and complex response patterns. With the rise of deep learning, generative adversarial nets provide a solution to missing data generation, in which GAIN is the most popular method.  However, there are three disadvantages of GAIN: first, it needs a hint mechanism to guarantee the model identifiability; second, it requires missing completed at random (MCAR); third, it does not consider the prediction after imputation. In this talk, we present a GAN based method to overcome these shortcomings and it shows promising theoretical and empirical results.