• 沒有找到結果。

Research and Writing Tips for


Academic year: 2022

Share "Research and Writing Tips for"


加載中.... (立即查看全文)



Research and Writing Tips for Graduate Students

Shou-de Lin (林守德) Professor

National Taiwan University sdlin@csie.ntu.edu.tw


Machine Discovery and Social Network Mining Lab, CSIE, NTU

PI: Shou-de Lin – B.S. in NTUEE – M.S. in EECS, UM

– M.S. in Computational Linguistics, USC – Ph.D. in CS, USC (EELD project)

– Postdoc in Los Alamos National Lab


– Social network Analysis

– Technical Writing and Research Method – Probabilistic Graphical Model

– Machine Discovery


– All-time ACM KDD Cup Champion (2008, 2010, 2011, 2012)

– Best Paper Award WI2003, TAAI 2010, and ASONAM 2011

– Google Research Award 2008 – Microsoft Research Award 2009 – IBM research award 2015

– INTEL research Funding 2011~2015

Machine Learning with Big


Discovery with Unlabeled


Learning in IoT Practical

Issues on ML/KDD



• Some of the materials and ideas are

originated from other people, including:

– Marie desJardins – Kevin Knight

– Ed Hovy

– Dianne O'Leary – Duane A. Bailey – Ronald T. Azuma

– Possibly others here and there



• How to find good research topics

• How to do good research

• Improving your RQ

• How to write a good paper


How to Find a Good Research



What is a Good Research Topic

• Something that interests you, your advisor, and your research community.

• A real problem, not a toy problem (or even worse, not a well-defined problem).

• Have certain connection to the existing research (If not, you need to make sure people think it is interesting

and worth doing.)

• There is a chance for you to have solid theoretical

contribution or practical/empirical results (preferably both).

• Significant yet manageable, with extensions and

additions that are successively riskier but will make the


A Good Research Topic Makes you Halfway to the Success

• A novel research topic is a big plus

– You know people would appreciate your work even before starting working on it.

– The topic itself is novel/interesting/challenging enough to have certain value.

– Usually you need to hurry up since somebody else might come up with similar ideas.

– Usually you need to do a lot of literature survey to make sure nobody does the same thing.

• If you cannot find a novel topic, then find a novel solution for an existing topic

– Sometimes the problem is trivial, but the solution is not.

– Novel solution is hard to come up with, so you don’t need to


What should I do if I cannot find a good research topic on my own

• Talk to your advisor and friends.

• Taking relevant courses.

• Read some papers.

• Don’t just read papers, do something (join a group, implement a system).

• Read tech news.

• Open yourself to new/novel/interesting ideas, even it has nothing to do with your expertise.


How to do Good Research?


Foundation, Foundation, Foundation

• Algorithm: Dynamic Programming, Graph theory, Clustering, automata, logic, cryptography

• Search methods:

– Optimization (e.g. heuristic search) : adjusting parameters of a system to optimize an explicit or implicit objection function (e.g.

Maximum likelihood Estimation)

– Learning (classification or regression): Given a set of

input/output pairs, learning tells you how to predict the output given some unseen input. Proposed methods: SVM, NN, ME, DT, GE, EM…

• Math: probability and statistics, information theory, coding theory, queuing theory, linear algebra, discrete math…

• Programming Skills: C++, Java, design related tools, Python, Perl, MPI, database management…

• Background knowledge in other areas: biology, music…


Finding your Own Hammers

• You need to identify your “secret weapon”.

• For example, the hammers in MS lab:

– Estimation-Maximization Algorithm.

– Master in classifiers (e.g. ME, SVM, DT, GA).

– Bayesian Inference Tools.

– Reinforcement Learning Packages.

– Probabilistic Graphical Model – Social Network Analysis Tools.

– Using Clustering Machines.

– Dealing with GigaWords of data


Find New and Better Ideas (knight)

• Listen to the data (Herb Simon)

• Kick around ideas with senior students and your advisor

– Reject mediocre ideas – Reject complex ideas

• Get animated by a giant goal

– Narrow it down immediately – what’s the first experiment?

• Learn powerful techniques by implementing

• Pick problems that will teach you somethingthem

• Obsess yourself with the research problem, and wait for the ideas to come.


Getting feedbacks

• To be successful at research, it is essential that you learn to cope with criticism, and even that you actively seek it out.

• Talking to other people will help you realize

– which aspects of your research are truly different and innovative – how your work fits into the current state of your field and where

it's going

– which aspects of your work are harder to sell (and, therefore, which aspects you need to think more about justifying).

• Give presentations at seminar series at your university, at conferences, and at other universities and research labs when you get the chance.

• Talk to people as much as they're willing to listen to. You should have 30-second, 2-minute, 5-minute and 10-minute summaries of your project ready at a moment's notice.


Improving your RQ (Research Quotient)

Unfortunately, IQ ↑ + EQ ↑ ≠ RQ↑



• Initiative

• Tenacity

• Discipline

• Flexibility

• Awareness

• Selective

• Ambitious but practical

• Get your hands dirty (mind the details)



• Rather aggressive than passive

– your adviser is NOT going to hold your

hands and tell you what to do every step of the way.

– Your goal is to prove that you can do high- quality research , not just to get a degree.



• "Let me tell you the secret that has led me to my goal. My strength lies solely in my

tenacity." - Louis Pasteur

– You don't need to be a genius to earn a degree, but very few finish a dissertation without being tenacious.

– No one can tell you in advance exactly how long the dissertation will take, so it's hard to see where the "end of the road" lies.



• Do research EVERYDAY, instead of doing it when you are in the mood.

• Try to find your own routine, and stick to it.

– Know which time slots in a day are best/worst for you mentally and physically.

• Simplify your life.

– Minimize distractions and detours.



• Working around problems if it is not possible to directly solve it

• Being willing to change plans if necessary

• Taking advantage of opportunities and synergies

• Accept the things you can't change (e.g. network broken).

– Control the controllables.

– Save the cursing time, it is YOU that should be responsible for how your time is spent.



• Pay attention to the rules, news, tips that benefit you.

• Be aware of the new opportunities (e.g. new research direction, new technology, new

scholarship, etc.)

• Have a sense or urgency. It is YOUR future.

• keep in touch with the "real world,“ remind yourself that the graduate student population is not representative of humanity in general and keep your own perspectives.



• Be aware that you have only limited amount of time (at most 24 hours a day).

• Don’t spend too much of your time on subordinate things or tasks.

– Learn how to say no


Ambitious but Practical

• Everything is possible, unless you prove it impossible.

• Don’t give up too early.

• Don’t settle for mediocre.

• Be realistic (there are more research to be done after graduation).


Get your hands dirty

• Walk the talk (Talk the talk, walk the walk).

– Smartness can be learned through experience.

• Have the determination to start working on a tough problem ASAP (Do it NOW!!)

• Knowing what is critical and what is minor (e.g. the speed is sometimes as important as the quality)

• The last 10% to perfection typically consumes 80% of the effort. The devil is always in the detail (Prof. Tzi- cker Chiueh)


How to Write a Good Paper


What’s so important about publishing?

• To convey your research ideas and results

accurately, clearly and economically to others.

• Presenting a coherent, written scientific argument is a learned skill – learn by doing!

• For application science, your works would not become applicable without first letting people know what it is.

• You’d be thrilled to realize that there are strangers reading your stuffs!

• To earn better understanding about your research

– Writing down your method usually can reveal its pitfalls.

– Sometimes it's difficult to define or formalize an idea well enough until you have written it down.


The Peer-reviewing Era

• You cannot publish your papers at will.

• You need to get the approval from a bunch of (usually anonymous) judges

– Good or bad?

• All of us have been reminded: a good writer should always consider the readers…

– It turns out your papers have two different kinds of readers: the reviewers and the normal readers

– The former determines whether your paper can be accepted, the later determines whether it will be cited


Why My Papers are Accepted/Rejected

Paper Acceptance?

Research Quality

Writing Skill Language skill

Others (reviewer,

luck, etc)

Good research  good paper

Good English  Nice paper writing


Research Quality: Meet Reviewers’ Demand

Writing Strategy: to show

(1) The applications. (1)The theoretical insight. Prove your results Provide detailed methods

Reviewers want


problems Novel Solutions Solid Results Repeatable outcomes


Things most likely to be criticized by reviewers

• Experiment (not enough, not convincing, not fair, no baseline, no confidence interval, etc)

• Methodology (too ad hoc, no complexity analysis, too complicated, too simple)

• Value (not important, not challenging, not applicable in the real world)

• References (too few, not citing somebody’s work)


Why My Papers are Accepted/Rejected

Paper Accept?

Research Quality

Writing Skill Language skill

Others (reviewer,

luck, etc)


Technical Writing Pyramid

4. Doing stage 1-3 elegantly.

3. Doing stage 1 and 2 efficiently (using as few

space as possible).

2. Presenting the methods and results in a convincing manner (that is, being able

to persuade others that you are doing great work).


How should one describe the idea

• The route you have come up with the idea.

• The way you implement your idea.

• The best logic you should present your idea.


What makes a good piece of technical writing

1. Clear and forthright

2. Convincing (logically sound) 3. Precise and familiar

4. Concise and fluid (smooth)


Clarity : what to avoid?

• A sentence that is too long

• A sentence that contains too many pronouns

• A sentence that contains too many relative pronouns (e.g, who, which, that)

• A sentence that contains many prepositional phrases (e.g, before the class, behind the door)

• A sentence that contains more than one idea.



• Writing down the most important points in this paper first.

• Prepare the skeleton (logical sequence of sections). Writing down section and subsection titles first.

• Write the introduction draft first, and go back to revise it after finishing the whole paper.

• Writing the abstract in the very end.

• For novice writers, a good strategy is imitation: choose a well-written paper that is of a similar flavor, analyze its organization, and sketch an organization for your results based on the same pattern.


Avoid grandiloquence

• Grandiloquence:

– The use of extravagant language – The use of long pompous words

– Creating a text that is difficult to read Example:

– It may seem reasonable to suggest that the necrotic effect may possibly due to toxins

Necrosis may be caused by toxins


Why My Papers are Accepted/Rejected

Paper Accept?

Research Quality

Writing Skill Language skill

Others (reviewer,

luck, etc)


The Impression to the Reviewers

• Avoid providing the following impressions to the reviewers:

– You are arrogant – You are a novice

– You are not confident about your research – You are not working as hard as you can


Be Humble but Confident

– Avoiding arrogant/exaggerate statements such as:

• We are the first team to …  To our knowledge, our idea of …is novel

– Be sure about your idea/proposal/results

• It seems to me that …  In my opinion, …


Don’t Feel Frustrated about Rejection

• Whether a paper gets accepted is sometimes a random process (i.e. the likelihood changes with time and environment)

• Sometimes the reviewers just cannot accept your submission, and it is easy to identify some flaws if they are determined to do so.


How to Improve Your

Writing/language Skills?

• Writing English papers and reports

• Writing more English papers and reports.

• Reading well-written papers (not necessary the best- known paper) from the writer’s point of view, and pondering:

– Why are they clear and easy to understand?

– The usage of language – The structure and flow

• Analyzing other people’s editing (why and how) on your write-ups.

• Don’t give up: Never feel that your writing skills cannot be improved (and don’t feel that it is all about English)



 When citing a foreword/introduction/preface/afterword, begin the citation with the name of the person who wrote it, then the word “Foreword” (or whatever it is), without

• School-based curriculum is enriched to allow for value addedness in the reading and writing performance of the students. • Students have a positive attitude and are interested and

3.16 Career-oriented studies provide courses alongside other school subjects and learning experiences in the senior secondary curriculum. They have been included in the

The Hilbert space of an orbifold field theory [6] is decomposed into twisted sectors H g , that are labelled by the conjugacy classes [g] of the orbifold group, in our case

OurChain stands for all your blockchains, an autonomous platform for any blockchain, including a ChainAgent, a ChainBrowser, a ChainFoudry, a Ch ainOracle and an OurCoin with

● In computer science, a data structure is a data organization, management, and storage format that enables efficient access and

Towards a model apprenticeship framework: A comparative analysis of national apprenticeship systems. Globalization and new industrial organization: Implication for

Acceptance of donations for naming rights (i.e. donations made on the condition that the receiving organization will name a specified “asset”, such as a building, a school,