Shulou(Shulou.com)06/03 Report--

Source: Xin Zhiyuan

This article is about 2300 words. It is recommended to read it for 6 minutes.

"regard learning as the ideal of life" Yao class begins a new journey!

Another young Chinese doctoral student in the direction of AI will soon join a prestigious American university to teach.

Wu Jiajun, a doctoral student from Tsinghua University and MIT, announced on Twitter that he will join Stanford University as an assistant professor of computer science in 2020.

Wu Jiajun said on Twitter: "I am pleased to tell you that I will join Stanford University as an assistant professor of computer science in 2020! thank everyone for helping me along the way. I look forward to the exciting research collaboration with Stanford AI Lab, Stanford Visual Learning Lab and Stanford people-oriented AI Research Institute."

Since the beginning of this year, apart from Wu Jiajun, three outstanding doctoral students in the field of AI have joined CMU, Princeton and other top universities, and their doctoral career is coming to an end. What awaits them is more challenging and meaningful scientific research work.

Chen Tianqi, author of XGBoost, will join CMU as an assistant professor in the autumn of 2020; Zhu Junyan, author of CycleGAN, will join CMU as assistant professor in autumn 2020; Jin Chi, winner of the best paper award for "RL Exploration" at the ICML 2018 seminar, will join Princeton as an assistant professor in September 2019.

Wu Jiajun himself is even more legendary. He graduated from Tsinghua University and is a 2010 (Program 00) student in Yao Class of Tsinghua Cross Information Research Institute.

Wu Jiajun

Wu Jiajun ranked first in the whole grade for three consecutive years during his undergraduate year, and won the first prize of Tsinghua University undergraduate Scholarship, Jiang Nanxiang Scholarship and Yao Scholarship.

Academically, Wu Jiajun has published a number of papers in the world's top conferences and journals, including three CVPR and one JAMIA, so that Yao Qizhi praised him for "his achievements in scientific research are comparable to those of doctoral students."

During the defense of the 2013 Tsinghua Special Award, Wu Jiajun's eye-catching transcript was exposed, so he was called "Tsinghua study God" by the majority of netizens.

After entering MIT, Wu Jiajun started the "open mode" of scientific research, including 2 CVPR, 2 ICCV, 4 NeurIPS in 17 years, 2 CVPR, 3 ECCV, 2 IROS, 5 NeurIPS in 18 years, and 1 TPAMI, 1 ICML, 3 ICRA, 8 ICLR so far in 19 years. The top of the major AI will be like his "back garden", the paper seems to be very simple.

However, "Genius + hard work" is the reason why Wu Jiajun has achieved such a success. when talking about how people think it is always easy for him to publish papers, Wu Jiajun once said, "I also have a lot of articles that are rejected and returned to make repeated changes before sending them out." it's just that people don't know. "

Opening and hanging in MIT: obsessed with Machine perception, 3D Vision and Robot Research

Wu Jiajun is currently studying as a doctoral student at MIT under professors Bill Freeman and Josh Tenenbaum.

On his home page, Wu Jiajun wrote that he studies machine perception, reasoning and the interaction between machines and the physical world, and draws inspiration from human cognition.

His research interests include:

Physical scene understanding dynamic model of robot manipulation visual reasoning to generate visual model multimodal perception

Academically, Wu Jiajun is very productive. Google scholar shows that he has published 59 papers, with a total of 2444 citations, with an h index of 21.

These are some of his recent academic activities and achievements:

Co-organized Vision Meets Cognition workshop and Sight and Sound workshop in CVPR 2019, and gave invited lectures on 3D Scene Generation, 3D-WiDGET workshop and 3D Scene Understanding workshop. Neuro-symbolic concept learner paper was selected in ICLR 2019 oral. "Learning to reconstruct shapes from invisible classes" was selected in NeurIPS 2018 oral. The paper "Let Robots learn to play stacked fun" published in Science Robotics, "combining physical simulators and neural networks" won the best paper award for IROS Cognitive Robot. Participated in organizing the NeurIPS'18 physical World Modeling Symposium and the RSS'18 Symposium on Imitation and causality.

From the above description, we can see that Wu Jiajun is very interested in the research on the combination of 3D vision, vision and cognition.

For example, in a 2018 paper, "MoSculp: interactive Visualization of shape and time", Wu Jiajun and his partners created an AI system called MoSculp, which only needs video input, uses AI to detect 2D images of the human body and restore them into 3D models, and can create surrealistic 3D motion sculptures.

Using AI to create 3D Sports Sculpture

They say the technique can be used to provide detailed sports research for athletes who want to improve their skills.

For another example, in the paper "See, feel, act: Hierarchical learning for complex manipulation skills with multisensory fusion" published in Science Robotics magazine in February, Wu Jiajun and others showed a robot that can play on top of each other:

They say the robot shows some capabilities that previous robotic systems could not: learn the best way to perform tasks quickly, using not only visual cues, but also tactile and physical interactions.

Yao Ban Xueshen: "I just do a little more than I asked."

Like many Yao students, Wu Jiajun was sent to Tsinghua through the National Informatics Olympic Games.

In September 2010, 18-year-old Wu Jiajun won the first prize in the National Youth Informatics Olympiad and was sent to the Institute of Cross Information of Tsinghua University. He was selected into the "Yao class" in the school's second enrollment examination, that is, the computer science experimental class of Tsinghua University.

Yao class master such as cloud, in the study, Wu Jiajun always strives for perfection. In addition to majored in computer science, he also studied for a double degree in economics and actively participated in time activities on and off campus.

"I believe that good academic achievement is not only the basis of scientific research work, but also the embodiment of learning attitude, and the down-to-earth and serious quality required to complete the course requirements is also fundamental to scientific research work." Wu Jiajun said this at the defense meeting of the special scholarship for undergraduates of Tsinghua University.

And his near-perfect grades proved that he did: he held the top grade in the three years of college, with a score of 94.0,96.6 and 98.0 respectively; a double-degree average of 95 in economics; and an An in all subjects during the second semester of his junior year at the Massachusetts Institute of Technology (MIT).

Wu Jiajun in the defense of Tsinghua Special Award

In response, Wu Chia-chun said: "I just do a little more than the teacher asked. I try to finish all the optional assignments and projects. I try to read all the bibliography provided, and if possible, read more books."

He said he chose a double degree in economics for the same reason: "on the one hand, it doesn't hurt to learn more, and on the other hand, it's because of interest in cross-disciplines."

In his sophomore year, he joined Microsoft Research Asia (MSRA) as an internship and officially opened the door to scientific research.

Initially, his research direction was natural language processing, and in cooperation with MSRA researchers, his "Medical record Co-reference Resolution system" won the first place in the Integrated Biology and Internal Information Challenge (I2b2 Challenge) held by MIT, and two papers were published one of which was published in the Journal of the American Medical Information Association (JAMIA), the number one journal in the field of medical informatics.

It was in MSRA that Wu Jiajun met Professor Tu Zhuowen of the University of California, San Diego (UCSD) and decided to focus on computer vision. Under the guidance of his tutor and his own painstaking study, Wu Jiajun has successively published one first author paper and two second author papers at the top conference on computer vision CVPR during his undergraduate course.

Papers published by Wu Jiajun during his undergraduate course

In 2014, Guangming Daily interviewed Wu Jiajun, talking about the future, he said: "I want to learn as a lifelong ideal."

At the time, the 22-year-old said, "if I have the opportunity, I want to enter the best academic circle in the world."

When asked about the ideal of "being the eye of China's computer science career and providing everyone with a better and happier life", Wu Jiajun expressed a rational understanding: "this is my ideal, but there is still a long way to go. I still have to study for many years."

About to join Stanford University to continue his AI research, Wu Jiajun is getting closer to the ideal step by step.

Editor: Yu Tengkai

Proofread: Tan Jiayao

-end-

Https://www.toutiao.com/a6727464139809096205/

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.