30-year-old second PhD, published 6 first-author papers in 17 months and won the best prize at the top conference! Real life restart simulator

Mengchen Yang Jing Bowen sent this from Aofei Temple
Quantum Bit Report | Public Account QbitAI

How do you evaluate a PhD graduate at the age of 34?

What if you are still someone who is determined to pursue scientific research and want to pursue a second doctorate degree?

Recently, such a doctoral student has caused heated discussions in the academic community, not because of his gains , but because of his sacrifices . A doctoral graduate who originally planned to pursue postdoctoral studies learned that his future research and teaching positions might be affected by the school rankings, and asked his postdoctoral "colleagues" a question:

What if I come here to study for a PhD ?

His name is Li Ang , and he is a student of Professor Chen Yiran at Duke University, researching federated learning.

Just a few days ago, his paper was included in SenSys, a conference on embedded network sensor systems hosted by ACM.

This is the sixth paper I have been accepted by a conference as the first author since April last year ! I also won the KDD 2020 Best Student Paper.

He was praised by Chen Yiran as a "rising star in system-level research on federated learning".

I am not an expert in system research. It is not so much that I taught him anything technically, but rather that he taught me for four years.

What is even more admirable is the inspirational study experience behind his crazy output.

He studied for his undergraduate degree at Henan University and then pursued a master's degree in software engineering at Peking University.

One year after graduation, he gave up his high-paying job in a bank and went to the University of Arkansas to pursue a PhD in computer science.

After that , I was admitted to the doctoral program for the second time by Professor Chen Yiran.

In his communication with QuantumBit, he emphasized:

It's never too late .

Because there is no so-called starting point and end point, the most important thing is to eventually achieve the goal you want.

Real life restart simulator

Li Ang had thought about doing research when he was an undergraduate student majoring in computer science at Henan University, and then he went to Peking University to study for a master's degree in software engineering.

After graduation, he worked as a management trainee in a bank in Beijing for a year, focusing on financial management.

Although working in a bank is very comfortable, Li Ang still feels that this kind of lifestyle is not what he really wants to pursue.

A year later, he decided to resign from the bank to pursue a doctorate and continue to pursue his academic ideals.

Li Ang received his first doctorate in computer science from the University of Arkansas, but computer science was not the school's strong point.

So he decided to continue his postdoctoral research, and his target was Professor Yiran Chen of Duke University, whom he had always been following.

Because his resume was not impressive enough, when he applied for Professor Chen Yiran's postdoctoral position, he was just trying it out and never thought that Professor Chen would respond to him.

△ The inspiration from Professor Chen Yiran’s postdoctoral recruitment

Unexpectedly, Teacher Chen not only contacted Li Ang, but also had a more in-depth communication with him.

Although now, Mr. Chen recalled and joked that it was because his alma mater, Henan University, was the same school as his parents, so he took one more look at him in the crowd.

But during that exchange, Li Ang felt that Teacher Chen was trying to understand a student's academic ideals and career plans from the bottom of his heart, rather than just judging a student on paper.

When Li Ang asked the question that changed his life, “What if I come to you to study for a doctorate?”, Chen Yiran’s answer was:

If you have the determination, everything is possible.

In the interview, Li Ang concluded that the internal reason for his decision to pursue a second doctorate was his own dedication and passion for academics, while the external reason was his admiration and worship for Professor Chen.

△ Professor Chen Yiran

At the age of 30, Li Ang is much older than other students in his group. However, it is not just his age that makes him anxious .

When he was studying for his first doctorate, Li Ang's main research direction was privacy protection and security issues related to mobile computing.

Regarding the AI-related research that Professor Chen is mainly doing, Li Ang feels that his knowledge accumulation in this area is not sufficient and he is not very familiar with the development of the AI ​​field.

At that time, he was very anxious and kept thinking about how to quickly find a breakthrough point and start research as soon as possible to catch up with the gap.

With this mentality, Li Ang went to his tutor Chen Yiran's office every day to talk to him about his ideas for the article.

Chen Yiran advised him not to rush to publish articles, but to start from learning research methods , finding problems, discovering problems, solving problems, and finally how to present the results in the form of papers.

This kind of training in a complete scientific research system is exactly what Li Ang lacked in his previous learning and research experience.

Later, Li Ang's research work at Duke University gradually got on track.

At the beginning, his papers were often rejected.

On the one hand, Li Ang began to doubt his own abilities, and on the other hand, he developed negative emotions towards the reviewers.

Chen Yiran used her own scientific research experience to enlighten him:

If you always treat the objective suggestions given by the reviewers with resistance, it will not help improve your research.

Starting from a paper submitted to Infocomm in 2019 that was rejected, Li Ang's mentality gradually changed.

He tried to think from the perspective of a reviewer: Was it his own research that was lacking, or was it the way the paper was presented that was lacking?

Following this idea, Li Ang and his team finally made major revisions to the article and added a lot of experimental data.

After the revision, I submitted it again to the MobiHoc 2020 conference and it was accepted smoothly.

This is also the first paper published by Li Ang as the first author at an international academic conference.

Later, Li Ang became more and more confident in himself and embarked on the "scientific research fast lane".

Since April 2020, his papers have been successively included in MobiHoc, KDD, IoTDI, MobiCom, SEC, and SenSys, and won the KDD 2020 Best Student Paper and SEC2019 Best Poster Awards.

Li Ang said that the so-called fast lane is only the result seen by the outside world from the output perspective. Behind it is a process of long-term accumulation and quantitative change to qualitative change.

During his second doctoral studies, he accumulated a lot of theoretical knowledge related to the field and read a lot of literature.

In addition to continuing to pay attention to papers in academic conferences and journals, I also keep an eye on the development trends in the industry.

His work experience in the bank enabled him to actively communicate with his seniors and classmates.

Of course, this is also inseparable from the guidance of Professor Chen Yiran, his support in purchasing various experimental equipment, and his encouragement to participate in international academic exchanges.

Most concerned about privacy computing

Having said so much, are you curious about what field Li Ang is researching that has attracted the attention of so many academic conferences?

Take this article that has just been included in SenSys, for example, it proposes a new federated learning framework

△ Source: From Li Ang’s Moments

This framework can reduce communication costs by 97% , improve model training and inference efficiency, and solve the problem of data heterogeneity in the federated learning training process.

It is divided into the following three aspects:

During the training process of federated learning, binary masks (rather than local model parameters) are passed. Binary
masks are learned using local data on the client, and data consistency issues are resolved by embedding personalized information of local data.
The parameters of the local model are frozen, and some structural sparsity constraints are added.

The topic of the paper, federated learning , is one of the main research directions of Dr. Li Ang.

He said that his overall research direction should be more applied systems, but the specific research content is still inseparable from theory.

For example, differential privacy technology requires very solid statistical theory knowledge as a foundation.

The challenges currently faced by federated learning: communication costs, data consistency, and insufficient computing power or performance, are precisely the issues he is mainly researching at this stage.

At the same time, finding better privacy protection technology has become Dr. Li Ang’s second research direction .

In this KDD 2020 best student paper, he proposed a data crowdsourcing framework for privacy protection: TIPRDC.

Experiments show that this is a framework that outperforms other existing methods in terms of utility-privacy trade-off.

Specifically, the TIPRDC framework can hide the private information in the intermediate representation while maximally retaining the original information embedded in the original data for the data collector to complete unknown learning tasks.

In other words, big data collection can be carried out more conveniently while protecting personal privacy.

Under this framework, larger-scale deep learning models and applications can be better trained and developed.

However, when applications are actually deployed, they are often limited by computing resources.

This brings us to Dr. Li Ang’s third research direction : developing a collaborative computing framework based on the edge/server side.

This paper at the edge computing conference SEC 2021 proposed a model called LotteryFL that is deployed on edge devices and has higher communication efficiency:

This framework can improve the inference accuracy by 17.24% and reduce the communication cost by 2.94%.

Looking through Dr. Li Ang's published papers, it is not difficult to see that privacy protection is his current biggest concern.

In an interview, Dr. Li Ang also said that in today's living environment where IoT devices are highly popular, the risk of privacy leakage we face is higher than ever before.

For example, virtual assistants, smart photography, and various smart devices that we see everywhere in our daily lives are collecting our personal data all the time.

However, most users have no idea where this data is stored, where it flows, and what its specific usage scenarios are.

How to ensure the transparency of data flow and provide users with clearer permissions to manage this data?

Dr. Li Ang believes that this is an issue that deserves further study.

It's never too late

Now Li Ang is about to graduate with his second doctorate degree.

He still insists on his original ambition to do research and plans to find a teaching position in academia.

When talking about his original intention of scientific research, Li Ang recalled his admiration for Steve Jobs when he first entered college.

I am particularly grateful for the power of constantly iterating smart devices and continuously advancing technology to truly change the world and human life.

Even when asked what book he would recommend, he mentioned Steve Jobs without hesitation.

This has become Li Ang's most original motivation and the ultimate goal of his research.

After gradually getting involved in research, Li Ang's love for scientific research has deepened: it can give me the greatest degree of freedom ,

On the one hand, I can expand my imagination and explore the cutting-edge areas that I want to learn about.

On the other hand, we can use cutting-edge technology to change our daily lives or integrate it into social production to solve problems.

It is precisely because of this reason, as well as the strong support of his family, colleagues and Professor Chen Yiran, that he can resist external criticism step by step and always firmly follow the path of scientific research.

Such an inspiring study experience has also inspired many people.

Some netizens said: Yes, if you are determined, you can actually do a lot of things!

Some people are suddenly interested in pursuing a PhD.

I also want to get a PhD at an advanced age.

Some netizens also felt the same way and thought about their own life experiences.

I have come a long way from a second-tier university, and I really have a yearning for scientific research. I have just started my second year of postdoctoral research.

But as Li Ang emphasized, it is never too late as long as you stick to your ideals.

He gave these suggestions to those who are determined to do scientific research.

First, don’t be anxious. You must calm down and do your own research steadily.

The second point is to find your true interest and clarify what you want. Once you have clarified your dreams and goals, you should start planning your future in advance, and then all that’s left is to work hard.

This is particularly critical, especially during the bottleneck period during doctoral studies, as your mindset will change many things.

In particular, Li Ang mentioned three essential abilities for students who are engaged in interdisciplinary research.

Quick learning, ability to discern professional research and the ability to communicate and collaborate.

Finally, I wish Dr. Li Ang all the best and may all his wishes come true!

Ps: Welcome scholars and teachers from the academic community to contact us at any time and communicate more ( ￣︶￣), contact email: ai@qbitai.com*

Home page:
https://sites.duke.edu/angli/

Reference link:
[1]https://weibo.com/2199733231/KxhNi2MXH

[2] https://www.linkedin.com/in/ang-li-3658273b/