Hacker News Reader

Comment by Darmani | Hacker News Reader

With all this talk of publication output, I'd like to see some examples of good research coming out of mainland Chinese universities.

In my field (programming languages), I can name one total paper that I liked that came from a mainland Chinese university. It stood out for that reason. A friend in another field (CS theory) complained similarly: massive amounts of shoddy work comes out of China, but nothing worth reading.

Meanwhile, China now has a spam journal called "Nature and Science," so you can tell your friends "I have 5 papers in Nature and Science."

View on HN

Ah, can I ask what paper it is? I’m in PL, and unless it’s coming out of MSR China, I haven’t seen PL papers that are interesting. China isn’t into PL, though, and it gets better in systems (big data, cloud, whatever is hot these days), there are some good labs with good output (Andy Yao’s algo lab at qinghua). Many of the researchers are foreigners however (just like in the USA).

I’m curious about the quality of published work coming out of china in ML. Has anyone who does ML research read recently a paper that came out of china that had something really useful to them in it? I think those are fairly good measures of output (a contribution found useful).

Hi Sean! We met a few weeks ago at OOPSLA.

The paper I have in mind is Zhai et al's "Automatic Model Generation from Documentation for Java API Functions" presented at ICSE last year, from Nanjing University. It is arguably more of an SE paper though, and a lot of the authors are at US/Canadian universities; for all I know, she did most of the work in the US.

Still, the number of good faculty who come from China is growing; it's only a matter of time before some of them choose to start research groups back home.

Mei Hong’s group among others do well enough at PKU in SE. Never saw anything out of NJU, interesting.

China’s normal research faculty system is really messed up, more so than the American system. Special labs like Andy Yao’s are actually funded well (so no need to body shop your grad students to make ends meet), but seem to be the exception. Let’s see what happens.

Body shop? Explain. How exactly is the system messed up.

It’s like you as a professor have grad students in your lab, not just a few like a western professor, but maybe 10 or even 20. A company needs some work done, you need the money, your grad students need to graduate, so...they use the lab to do the work for the company, professor gets paid (important since their normal salary is 10k RMB/month or less), grad students get to graduate, company gets cheap labor....

It was really annoying to us because the professors would always keep their best students (especially undergrads, the good ones were better than grad students as they would go abroad and not become grad students in china) not allowing them to work for us as interns unless we had really tight guanxi with them. America is still 10 times better than china in this regards. Unless they root out the corruption, they are never going to advance beyond paper farms no matter how much money they throw at it.

Interesting. So, I understand that the grad students are not interested in the research. Are they writing papers on these "projects?" What good is a Chinese Ph. D. then?

Some European (Belgian?) students told me that, in their country, it was common to have an "industry-funded" Ph. D. (hence all the people I've met doing DSLs for insurance companies). I wonder to what extent their system is similar.

Think of it like a very focused internship. You still get something out of it (you did the work after all!), but it’s not like you got to explore your own thing and there was a lot more grunt work involved; since the professor had 20 student you also got a lot of experience in managing other students (senior PhDs will direct junior PhDs and master students, who will direct undergrads if any, you might rarely see the professor unless you are at the top of the lab). In that sense, many employers might find a chinese grad degree more preferable to a western one. On the other hand, research breakthroughs are much less likely.

I think China is huge in Computer Vision. And in terms of good Chinese papers ... there are really a lot of them in this Area, some of them are hugely influential.

Example? (Of very influential Chinese papers in computer vision)

https://arxiv.org/abs/1512.03385 Deep Residual Learning for Image Recognition Citation 4000+. Milestone of image recognition.

https://arxiv.org/abs/1502.01852 Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification. Also known as PreLU, citation 1600+. The initialization scheme is widely adapted by a lot of following works.

Those are the most influential ones, all come from Kaiming He. He is now in FAIR, and keeps very good publication record, with works like MASK-RCNN, but that doesn't really count as purely Chinese contribution, so we skipped here.

Some recent memory: 1. https://arxiv.org/abs/1703.06211

2. https://arxiv.org/abs/1709.01507

3. https://arxiv.org/abs/1707.01083

4. https://arxiv.org/pdf/1608.06993

5. https://arxiv.org/abs/1705.01088

Those are all good papers.

We asked about papers from universities in mainland China. You posted 7 papers. One of them had a second author from Shanghai Jiao Tong, and another had a second author from Tsinghua. The rest were from industry, MSR, and western universities.

So, we had someone from another field go searching for examples of good research from mainland Chinese universities, and come up empty-handed. I think this counts as further evidence for mainland Chinese universities being paper farms. :)

And, what do you know? Kaiming He did his Ph. D. in Hong Kong, not mainland China.

You might not take a close look, the one from Tsinghua is not second author, it is noted as an equal contributor. And there may be two western universities in total, taking into account of all authors. I like how your logics goes, that it is up to you to assign the credit as how much goes to who, everything doesn't fit your narrative doesn't count. As to your trivia, I don't think it matters, since his best research is done in MSRA, with coauthors that are mainland Chinese, not in that Hong Kong University.

Majority of authors of the papers are from China, the research is done by Chinese, mainly in mainland China, I believe they are Chinese enough, the point I want to make is obvious. Otherwise, I feel majority of the US CV papers would be disqualified because they are likely not done by locals here.

Last but not least, I was answering to the question as regards to good Chinese papers, not specifically from Chinese universities, not directly towards your question after all.

Sean worked at Microsoft China; I don't think he's worried about Chinese output. There was never any question that lots of people from mainland China do great research (otherwise my department is in trouble). The question is about universities in mainland China. This is kinda the equivalent of me saying that the City University of New York isn't very good, and then you trying to name smart people from New York to disprove it.

I understand you may be upset that I didn't like your answer, but I ask that you assume good faith and not accuse me (us?) of believing something we didn't say.

I think it is then easy to settle. I wasn't answering your question, my answer is never about Chinese university anyway.

If u feel being accused of and it is not what u mean, I am sorry. However, your post comes with a fighting spirit, and a trollish account, that seems to designatedly trigger people, if you expect good faith, I would suggest u to improve on the use of language

Apology accepted. I do see why my posts would provoke a reaction in someone who confuses "Chinese universities" with "Chinese people." If that's not you, then I'm mystified.

What you observed is exact how we set the goal and play the game. As a Chinese, I think your observation is very insightful to point to the model/structure of these kind of advance measured by normal ways. But most people follow the normal fact checking measurement. That's one of the powerful way to manipulate people's perspectives by improving on some facts. In the end of the day, we all react to some signals.

Reply on news.ycombinator.com