How Zhou Zhihua, a Purely Domestic Leader, Took up the Banner of the Intelligent Academic Community

IJCAI 2019 is in progress. As the most anticipated special report session of the conference, among the 7 heavyweight guests this time, there is Professor Zhou Zhihua, who we are familiar with. He gave a keynote speech on deep learning this afternoon. Take this opportunity to watch the sharing of the great god at the conference, and deeply understand his footprints on the road of AI exploration, and fully restore his contribution and promotion of AI development.

As a top scholar in the field of AI, Professor Zhou Zhihua has devoted great enthusiasm to scientific research, talent cultivation, and publication dissemination, and his extraordinary academic contributions have earned him numerous awards.

At the top conference IJCAI 2019, Zhou Zhihua also attended as scheduled. What did he share this time? What other interesting stories are there behind this star scientist?

What did Zhou Zhihua share at IJCAI?

IJCAI 2019 is halfway through, and many highlights have emerged. Professor Zhou Zhihua, as a distinguished guest, gave aInvited Talks), which attracted the attention of the whole audience.

In the report that started at 2 p.m., Zhou Zhihua delivered "Deep Learning: Why deep and is it only doable for neural networks?"speech.

He first introduced the connection between deep learning and deep neural networks (DNNs), reviewed the development history of deep neural networks, and pointed out the fact that the number of network layers is constantly increasing.

However, as the number of layers of deep neural networks increases, some shortcomings are also exposed, such as easy overfitting, complex training process, and increased computing costs.

This raises the question: Is deep learning equivalent to DNN (can deep models only be constructed through differentiable function structures)?

Professor Zhou Zhihua then pointed out another solution for deep learning and introduced the technology they were researching. gcForest (multi-Grained Cascade Forest),Drawing on the feature representation learning and ensemble learning of DNN, we use the decision tree forest method to achieve the effect of a deep model.

It consists of two parts: Cascade Forest and Multi-grained By introducing and comparing with other models, it is claimed that this is the first deep model that does not require neural networks and BP algorithms to implement. At the same time, it lists the actual use cases in which gcForest actually performs better than other models.

He then looked forward to the future of deep learning, pointing out the difficulties that may be faced if the development relies entirely on deep neural networks, and that gcForest abandoning neural networks would be an effective attempt.

However, he also emphasized that this technology is still under exploration and has a long way to go before it matures, and then he specified the challenges faced by the model. He also looked forward to the environments where deep forests and DNNs excel respectively, as well as the possibility of combining them in the future.

In addition to this special sharing full of practical information, he will also serve as IJCAI 2021 The processChairman of the Order Committee,This is the first scholar from the mainland to receive this honor.

He once said in an interview:It's not me who is awesome, it's the Chinese community that is awesome. This reflects that the strength of Chinese scholars has been recognized.I hope that everyone will work together to further enhance the contribution and influence of Chinese scholars and continue to produce excellent research results. "

So how did such a great person reach the pinnacle of machine learning?

Local AI master: Interest is the best teacher

In 1995, when Zhou Zhihua was a junior in college, he found Machine Learning: The Path to Artificial Intelligence in the library, which introduced him to machine learning for the first time. The following year, Zhou Zhihua graduated from Nanjing University at the age of 21 and chose to stay on to continue his studies.

After a period of exploration, he was attracted by theMachine Learning and Data MiningAttracted by this, driven by his interest and his own unremitting efforts, he made rapid progress on the AI path.

He graduated two years early during both his master's and doctoral studies. He stayed at NTU to teach in 2001, and was promoted to associate professor the following year. A year later, he was appointed as a professor, and subsequently obtained the qualification of doctoral supervisor. In 2006, he was selected as a Yangtze River Scholar by the Ministry of Education...

When asked about this amazing resume and the topic of talent, Zhou Zhihua said,"Interest is the best teacher"He also quoted Marx's words, "When labor becomes man's first need, rather than a means of making a living," to express the importance of interest.

This "local scholar" who completed his studies entirely at NTU has also received unanimous recognition from the international academic community.

From 2012 to 2016, Zhou Zhihua successively became a"Fellow" of the five major international associations: ACM, AAAI, AAAS, IEEE, and IAPR.

Became an International Society for AI The first Chinese Fellow to achieve the Grand Slam.In 2017, Zhou Zhihua was also electedForeign member of the European Academy of Sciences.

In addition, he has served as chairman of dozens of international academic conferences, including the International Conference on Artificial Intelligence held some time ago. (AAAI 2019), who also served as Chair of the Program Committee.

Over the years, he has published papers in leading international journals and top international conferences. More than 200 articles,Obtained invention patent 20  Remaining items.

In a recent statistics released by the Acemap team of Shanghai Jiao Tong University, Professor Zhou Zhihua ranked first among Chinese scholars in the ranking of the influence of AI scholars' papers in the past decade.Fifth place,The number of citations reached 4452.

Zhou Zhihua and AI education in China

In addition to his many academic achievements, Professor Zhou Zhihua is also devoted to the cultivation of AI talents in China.

In 2004, under his leadership, Nanjing University was established  LAMDA (Learning And Mining from Data) Research Institute, focusing on machine learning and data mining.

After more than ten years of hard work, LAMDA has become a leading research group in China. Students interested in the field of machine learning call it a "holy land", and a large number of students apply every year.

There are also letters circulating on the Internet that Teacher Zhihua wrote to applicants, which help applicants analyze in great detail from a teacher's perspective whether they are suitable for LAMDA.

Last year, under the leadership of Zhou Zhihua and others, NTU establishedSchool of Artificial Intelligence,AI undergraduate education was launched, and Zhou Zhihua himself became the dean. This was undoubtedly a shot in the arm at a time when AI talent and education were both scarce.

Facing various voices from the outside world and the industry, he is also very determined to insist on becoming an independent institute. In a media interview, he said,"Instead of patching up the existing framework, it is better to build it from scratch based on the characteristics of the artificial intelligence discipline itself."

The School of Artificial Intelligence of Nanjing University opened its first undergraduate enrollment in September 2018, and Zhou Zhihua will personally teach. When talking about the purpose of the School of Artificial Intelligence of Nanjing University, he revealed that they had started to study the undergraduate education program very early.

He said that the college is not aiming to form a training course like "Artificial Intelligence in 120 Days from Beginner to Mastery", but to cultivate high-level talents who can innovate in the field of AI and solve key technical problems.

The most classic reading material in the field: Watermelon Book

In fact, when it comes to Zhou Zhihua, many people are more familiar with Machine LearningThis book. He uses humorous and popular language to run through the key points and difficulties of machine learning. The book also uses watermelons as examples to make the theoretical knowledge vivid, so it is also called the Watermelon Book.

Xiguashu has been published for many times since its release in 2016. It is also often active on the bestseller lists of major e-commerce platforms, opening the door to machine learning for countless learners.

He once revealed the two purposes of this book in a conversation:

First, for beginners to understand the field of machine learningDraw the outline of the whole.Know what is available in this field so that you can further study it according to your needs and interests.

Another is to facilitate readers to read a variety of materials,Take a look at the knowledge system.

Interestingly, in the 2017 "Jingdong Literature Award" selection, Xiguashu was ranked as the annual new work award. TOP 5 . 

The book begins like this. Let's experience Zhou Zhihua's writing skills together:

In the evening, the road surface of the small street was moist after the light rain, and the warm west wind blew. I looked up at the sunset in the sky. Well, tomorrow will be another good day.

I walked to the fruit stall and picked up a green watermelon with a curled root and a loud gurgling sound. I was looking forward to the refreshing feeling of eating the thin skin, thick flesh and sweet meat. At the same time, I thought happily that I had worked very hard this semester, had a clear understanding of the basic concepts, and could handle algorithm homework with ease. I would definitely get a good grade in this course!

Let's look forward to more awesome research from the great god

As a great man in the field of artificial intelligence, Zhou Zhihua is undoubtedly a leader in the industry. However, he has said on many occasions that his greatest pleasure is scientific research and his greatest wish is that his research group can makeBetter research,Internationally there areMore impactful work.

When asked how he chose machine learning when AI was not valued, his answer was simple but powerful:"Stick to your own judgment and do important things."

I wish Mr. Zhou Zhihua can realize his wish and produce more and better results.