Pan Ling: PhD Study Is a Journey of Both Hardship and Joy

In the Young Authors Roundtable Forum at this year's IJCAI-SAIA YES conference, hosted by Associate Professor Yan Junchi from the Department of Computer Science and Engineering of Shanghai Jiao Tong University, six first authors of IJCAI 2020 selected papers were invited.

Among them, Pan Ling became the focus of everyone's attention.

The top student in her major for four consecutive years

Pan Ling attracted attention not only because she was the only girl in this session, but also because of her past academic achievements.

In 2013, Pan Ling entered the School of Data Science and Computer Science at Sun Yat-sen University, majoring in software engineering.During my undergraduate studies, I ranked first for four consecutive years (a total of 435 people) and received the National Scholarship for three consecutive years.

In the second semester of her junior year, she was successfully admitted to the Institute of Cross-Disciplinary Information Sciences at Tsinghua University and was given the opportunity to pursue a direct doctorate degree under the tutelage of Associate Professor Huang Longbo.

After completing her PhD, Pan Ling interned at the Machine Learning Group at Microsoft Research Asia and visited the research group of Professor Shimon Whiteson at Oxford University.He has won the Microsoft Scholarship, the Tsinghua University First Prize Scholarship, the Institute of Cross-Disciplinary Information Sciences First Prize Scholarship, etc.

On the road of scientific research, choice is more important than persistence

Currently, Pan Ling focuses on scientific research in deep reinforcement learning, multi-agent systems, and computational sustainability.

So far, she has published several first-author papers, which have been included in top conferences such as NeurIPS, AAAI, IJCAI, and AAMAS. Among them, her article in AAMAS 2020 (previous 5%) was invited to be submitted to the fast-track review channel of JAAMAS. In addition, Pan Ling has also served as a reviewer for conferences such as ICML, NeurIPS, and ICLR in recent years.

Before receiving these honors, Pan Ling was confused about whether to continue on the path of computer science.

Before college, Pan Ling had always been interested in economics and management majors, but by chance she entered the School of Data Science and Computer Science at Sun Yat-sen University.

In the second semester of her freshman year, she was still hesitating whether to continue studying computer science and was even preparing for a major change exam. It was not until she had in-depth discussions with her teachers and reflected on herself that she gained a new understanding of her major and her future.It also strengthened her determination to continue studying computer science.

In 2016, AlphaGo defeated the Go king Lee Sedol 4:1. This battle shocked the world and made deep reinforcement learning known to everyone.

In the same year, Pan Ling, then a junior in college, entered the stage of choosing a school for graduate school admission. She chose to study directly for a Ph.D. at the Institute of Cross-Disciplinary Information Sciences at Tsinghua University.

At the suggestion of her supervisor, Associate Professor Huang Longbo, Pan Ling read papers on reinforcement learning for the first time.

Pan Ling talked about her understanding of reinforcement learning.She believes that reinforcement learning can model and learn many scenarios of timing problems in real life.

For example, how we learned to walk when we were children is a very typical example. In different states, we should know what behaviors we should take to get rewards, that is, to walk successfully without falling.

Reinforcement learning is a very general framework for solving real-world problems.Therefore, Pan Ling believes that reinforcement learning has great application value and there are many areas worth exploring.

Paper rejected five times? Don’t lose your mind

In Pan Ling's scientific research career, she has also encountered repeated rejections of her papers.Pan Ling's first-author paper "Reinforcement Learning with Dynamic Boltzmann Softmax Updates" was rejected five times and was finally selected (the acceptance rate was 12.6%).

According to Pan Ling, the research direction of the paper is the accuracy of value function estimation, which is an important task in reinforcement learning. The research and writing process of the paper went smoothly, but after the paper was completed, it was repeatedly rejected and revised.It took nearly two years to finally get the paper included.

The smooth start of the proposal and the first draft, and the lengthy review and revision process brought a big gap to Pan Ling, which also gave her time to reflect.

Regarding the problem of "collision of topics" in scientific research, she feels that on the one hand, more detailed research on the research direction is needed, but if collision of topics is inevitable, there is no need to be discouraged.The reason for the overlap of topics is that this research direction is recognized by more people, and popular directions will also produce more valuable applications.

There are many solutions to every problem.We need to make a more detailed comparison of previous methods and strive to find innovative and influential approaches.

In the face of the "involution" phenomenon that is hotly debated by the whole society, Pan Ling also put forward her own views. In the current academic circle, different people will have different pursuits for the number and quality of publications.

In order to continue on the academic path, Pan Ling has clear academic quality requirements for herself. She will not make too many comparisons with others, but hopes to have more valuable explorations in reinforcement learning to help reinforcement learning truly take off.

Tips from a senior

When we asked what qualities young scholars should possess, Pan Ling emphasized rigor and responsibility.She believes that scientific research requires absolute rigor and professionalism, and she is responsible for the work she completes.

Academics is not an easy path. At the beginning, we may have great ambitions and want to achieve something, but the process will definitely not be as easy as we imagined.

When you are stuck or your paper is rejected, you should not be discouraged by the pressure. At the same time, you should stick to your original academic intention when facing temptation.The four most important words are don’t give up.

Pan Ling herself would follow the developments of the big names in her field through Twitter.Get the latest progress of reinforcement learning in time. At the same time, you can exercise by following Pamela's videos to relieve stress and balance your body and mind.

At the end of the interview,Pan Ling mentioned that she has always liked Marvel movies.In every movie, superheroes often suffer a lot of hardships, but in the end they are always able to turn the tide and save the world.

It is the brave and tenacious Marvel heroes who encouraged Pan Ling to maintain confidence, passion and persistence during her doctoral studies, and to continue to walk firmly and confidently on her future academic path.