In 2020, Songchun Zhu returned to China as a strategic scientist of artificial intelligence, and established the Beijing Institute of Artificial Intelligence, and served as the president. At the same time, he served as Professor Peking University, Dean of Peking University Institute of Intelligence, Dean of Institute of Artificial Intelligence and Professor of Basic Science in Tsinghua University.
Songchun Zhu 199 1 Graduated from China University of Science and Technology, majoring in computer science; 1992 went to study in the United States, and 1996 received a doctorate in computer science from Harvard University. In the same year, he entered the Department of Applied Mathematics of Brown University in the United States for postdoctoral research.
2065 438+00-2020, served as the chief scientist of MURI, an interdisciplinary cooperation project in the fields of vision, cognitive science and AI in the United States; In 2020, he returned to China to build the Beijing General Research Institute of Artificial Intelligence, and at the same time served as Professor Peking University, Professor of Basic Science in Tsinghua University and Dean of Peking University Institute of Artificial Intelligence; 202 1, Peking University intelligent research institute was established and served as the president.
Songchun Zhu's main achievements:
1995 From 2005, Professor Songchun Zhu, his tutor Mountford, colleagues and doctoral students at UCLA established a unified mathematical model for the concept of [early vision] proposed by David Marr, the founder of computational vision, including [texture], image primitives and primitive sketches.
The minimum and maximum entropy principles of statistical modeling are put forward. The findings of neurology and psychology are implanted into the [Gibbs model of statistical physics, thus a new probability model [framework] of Markov random field is derived, and this model is extended to the intermediate visual model to describe the principle of shape and gestalt composition.
Then, the statistical laws of scale invariance and scale change of natural images are found, and various visual modes and their corresponding mathematical models are mapped to a continuous entropy spectrum and information scale; Then, the mechanism of jumping and perceptual transformation between various models is studied, and the theory of perceptual scale space is deduced with doctoral student Wang Yizhou.