Lei Xu

Timeline

1. By 1986

   Harbin and Tsinghua degrees

   He earns a bachelor's degree at the Harbin Institute of Technology, then master's and doctoral degrees at Tsinghua University, where he completes his PhD thesis by the end of 1986.

2. 1987–1993

   Peking University and MIT

   He becomes a postdoc in Peking University's Department of Mathematics in 1987 and, in 1988, one of ten young researchers there given an exceptional promotion to associate professor. In 1989, he and Judea Pearl publish a chapter on structuring causal tree models with continuous variables in Uncertainty in Artificial Intelligence 3. From 1989 to 1993, he works as a postdoc and senior research associate in Finland, Canada and the United States, including Alan Yuille's team at Harvard and Michael Jordan's team at MIT.

3. 1990–1993

   Randomized Hough Transform and RPCL

   In 1990, he and co-authors including Erkki Oja publish the Randomized Hough Transform (RHT) as a new curve detection method in Pattern Recognition Letters. A 1992 paper with Adam Krzyzak and Ching Suen sets out several methods for combining multiple classifiers for handwritten character recognition. In 1993, he publishes Least MSE Reconstruction (LMSER) as a principle for self-organizing nets and, with Krzyzak and Oja, Rival Penalized Competitive Learning (RPCL) for clustering analysis, RBF nets and curve detection.

4. 1993–2016

   Chair professor at CUHK

   He joins the Department of Computer Science and Engineering at the Chinese University of Hong Kong as a senior lecturer in September 1993 and becomes a professor in October 1996. He serves as a chair professor there from 2002 to 2016.

5. 1995–1998

   Mixtures of experts and BYY learning

   In 1995, he coauthors "An Alternative Model for Mixtures of Experts" with Michael Jordan and Geoffrey Hinton, and in 1996 he and Jordan publish convergence properties of the EM algorithm for Gaussian mixtures in Neural Computation. In a March 1998 guest lecture, he presents his Bayesian Ying-Yang (BYY) system and theory, a unified statistical learning approach for supervised learning, unsupervised learning and temporal modeling that also gives new criteria for choices such as the number of clusters or hidden units.

6. 1995–2006

   IEEE and IAPR fellowships

   The International Neural Networks Society gives him its Leadership Award in 1995, and he is elected an IEEE Fellow in 2001. The International Association for Pattern Recognition names him a Fellow in 2002 for contributions to studies on the Randomized Hough Transform and combination of multiple classifiers, and the European Academy of Sciences elects him in 2003. In 2006, he receives the Asia Pacific Neural Network Society's Outstanding Achievement Award, after Amari in 2004 and Fukushima in 2005.

7. 2016–2019

   Zhiyuan chair at Shanghai Jiao Tong

   He joins Shanghai Jiao Tong University as a Zhiyuan chair professor in summer 2016, and in December 2016 five teams from four departments of its School of Electronic Information and Electrical Engineering form the Center for Cognitive Machines and Computational Health (CMaCH) with him as director. The center opens formally in June 2017. In October 2019, as chief scientist of the SJTU AI Research Institute, he lectures at the European Academy of Sciences' symposium on artificial intelligence about deep bidirectional learning, from Lmser reconstruction and BYY harmony learning to variational autoencoders, GANs and U-net.

8. As of October 2026

   Emeritus professor at CUHK

   In February 2026, he co-authors PAIRNet in PLOS Computational Biology from SJTU's Center for Cognitive Machines and Computational Health; the deep learning framework predicts PIWI-mediated RNA cleavage rates. He is an Emeritus Professor of Computer Science and Engineering at the Chinese University of Hong Kong.