How to do research and How to write paper By Zhi-Hua Zhou

How to Do Research & How to Write a Paper
周志华 2001. 5. 19
Copyright ? 2001 Zhi-Hua Zhou
Why do research?
?Curiosity ?Investment for future ?Promotion ?Contribute to Sci. & Technol. ?……
Copyright ? 2001 Zhi-Hua Zhou

Why write papers?
?Make your work recognized ?Provide basis for further research ?Promote your professional reputation ?…… Good research is the soul of a good paper
Copyright ? 2001 Zhi-Hua Zhou
做研究的总决 ?昨夜西风凋碧树 独上高楼 望尽 天涯路 ?衣带渐宽终不悔 为伊消得人憔悴 ?众里寻他百千度 蓦然回首 那人 却在灯火阑珊处
Copyright ? 2001 Zhi-Hua Zhou

How to do research?
?Choose a good topic ?Master the state of the art ?Good idea ?Sound supports ?Reasonable explanations ?……
Copyright ? 2001 Zhi-Hua Zhou
What is a good topic?
?You have interests in ?Potentially useful ?Not outdated ?Neither too big nor too trival ?Manageable to you ?……
Copyright ? 2001 Zhi-Hua Zhou

How about HOT topics?
a few significant works much HOT classic or meat significant works full of classic meat works whole crop HOT in China
few significant works mixture of meat and bone bone trival works
common researchers laggards pioneers leading experts novices
Copyright ? 2001 Zhi-Hua Zhou
?Try to initialize HOTs ?Try to catch future HOTs ?Try to follow international HOTs ?Care to follow Chinese HOTs ?Never to be laggards
Copyright ? 2001 Zhi-Hua Zhou

How to master the state of the art?
?Query your advisor or senior researchers ?International communication
attend conferences, personal communications
?Reading
about 20 papers for a new area
?……
Copyright ? 2001 Zhi-Hua Zhou
What to read?
?Books by leading experts ?Tutorials ?Good surveys ?Classic papers ?Very related papers
Copyright ? 2001 Zhi-Hua Zhou

How to identify classic papers?
?Query your advisor or senior researchers ?Internet
some area has its own URL
?References of papers
more cited, more classic
?…… Don’t ignore the references
Copyright ? 2001 Zhi-Hua Zhou
How to get reading materials?
?NJU library
Elsevier, Kluwer, Springer
?Internet
author’s URL, Institution’s URL, search engine, mailing lists, ……
?Email the author ?Your friends abroad ?……
Copyright ? 2001 Zhi-Hua Zhou

What is a good idea?
?Original ?Useful ?Reasonable ?Feasible ?A real progress instead of trival improvement ?……
Copyright ? 2001 Zhi-Hua Zhou
A good idea is the core of good research
the contributor of the idea usually acts as the first author
Never steal other’s ideas !
Copyright ? 2001 Zhi-Hua Zhou

When you get an idea, Don’t realize it immediately Instead, you should
?Check whether the work has already been done by others ?Ask yourself: Is it feasible?
Bad idea will kill life
?Lay the idea aside, come back a month later and ask yourself: ?Are there better topics? ?Are there better ideas? ?Is the idea feasible? If all the answers are “yes”, then realize the idea a.s.a.p. In this way, the chances of failure will be greatly lowered
Copyright ? 2001 Zhi-Hua Zhou
What are sound supports?
?Solid theoretic analysis ?Solid experiments
Well-defined experiments Benchmark tests Data sets available to other researchers Comparisons Repeatable results
Copyright ? 2001 Zhi-Hua Zhou

Why explanations?
No work is perfect ?Strength and weakness
How/when strong Why strong How/when weak Why weak
Copyright ? 2001 Zhi-Hua Zhou
How to write a paper?
?Choose an appropriate target ?Misconceptions ?Paper style ?Tricks ?……
Copyright ? 2001 Zhi-Hua Zhou

How to choose a target?
Know the hierarchy of publications in your area
Copyright ? 2001 Zhi-Hua Zhou
top international journals
such as Artificial Intelligence, Machine Learning most are milestone works
first-class international journals
such as JAIR, IEEE Transactions most are significant works
second-class international journals
such as most journals of Elsevier, Kluwer, some of Springer most are very good works
third-class international journals
most international journals most are good works
Other journals
Copyright ? 2001 Zhi-Hua Zhou

top conferences
such as IJCAI comparable to first-class international journals
first-class conferences
most ACM Conferences, Some IEEE Conferences, etc. comparable to second-class international journals
second-class conferences
most IEEE Conferences, etc. comparable to third-class international journals
third-class conferences
most international conferences held in China
Copyright ? 2001 Zhi-Hua Zhou
top native journals
中国科学 科学通报 comparable to second-class international journals
first-class native journals
计算机学报 软件学报 电子学报 自动化学报 etc. comparable to third-class international journals
second-class native journals
计算机研究与发展 模式识别与人工智能 …… comparable to third-class international conferences
third-class native journals
计算机科学 小型微型计算机系统 etc.
Other journals
most Chinese journals
Copyright ? 2001 Zhi-Hua Zhou

Journals are archival publications
Don’t write two journal papers with a same idea! Invited, regular, correspondence usually about 1/2 proportions are experiments
Conferences are intended for communication
Papers can be expanded to journal papers Keynote, Plenary, oral, poster fewer experiments than journal papers
Don’t care SCI & EI, to be cited is more important !
Copyright ? 2001 Zhi-Hua Zhou
Great work
submit to top international journals, top conferences, firstclass international journals
Very good work
submit to second-class international journals, first-class conferences, top native journals
Good work
submit to third-class international journals, second-class conferences, first-class native journals
Moreover, ……
Copyright ? 2001 Zhi-Hua Zhou

Knowing the review criterion may be helpful most papers will be reviewed by at least 3 reviewers that are experts in your area
best papers: very good papers: good papers: potential good papers: accept without change accept with minor change accept with major change about 1% about 3% about 6%
reject with resubmission encouragement about 25%
about 15% papers will be rejected without review Remember that most competing papers are good ones
Copyright ? 2001 Zhi-Hua Zhou
review form 1
a famous international journal
Copyright ? 2001 Zhi-Hua Zhou

review form 2
a top conference
Copyright ? 2001 Zhi-Hua Zhou
review form 3
an international journal
Copyright ? 2001 Zhi-Hua Zhou

review form 4
a first-class native journal
Copyright ? 2001 Zhi-Hua Zhou
In summary, reviewers will check:
?Does the paper introduce a new problem or provide a new solution to an existing one? ?What is the main result of this paper? ?Is the result significant? ?Is the paper technically sound? ?Does the paper provide an assessment of the strength and weakness of the results? ?Is the paper clearly written so as to accessible by most researchers in this area? ?Does the paper refer appropriate related works? ?……
Don’t fool the reviewers !
Copyright ? 2001 Zhi-Hua Zhou

Misconceptions 1:
The more, the better Don’t provide too much “new” ideas in a paper. Try your best to focus the readers on your core work.
Copyright ? 2001 Zhi-Hua Zhou
Misconceptions 2:
The more complex, the better Try to make your paper easy to be understood. Remember: You are showing your thoughts instead of confusing the readers
Mathematical language is to help you clearly express your thoughts when natural languages cannot
Copyright ? 2001 Zhi-Hua Zhou

Misconceptions 3:
The more selling, the better Don’t exaggerate too much on your work. The readers will judge it for you.
Copyright ? 2001 Zhi-Hua Zhou
Misconceptions 4:
The more authoritative, the better Don’t refer too much own works while ignoring others.
Copyright ? 2001 Zhi-Hua Zhou

Paper style
?Outline ?A template ?……
Copyright ? 2001 Zhi-Hua Zhou
Outline
?Problem X is important ?Previous works A, B, and C have been done ?A, B, and C have their weakness ?Your work D ?(Theoretic analysis of D) ?Comparison against A, B, and C ?Why D is better ?Strength and weakness of D ?Future works on D
Copyright ? 2001 Zhi-Hua Zhou