Advice on Creative Thinking, Research, Writing and Speaking
A collection of advice about how to do research and how to communicate
The educational objectives of a Ph.D. program
Pursuing Ph.D. is to develop the following skills:
- Life-long self-learning capability
- Technologies change very quickly. Moreover, in your whole career
life, you may need to change your technical areas multiple times. Hence,
you need to be able to learn new technologies by yourself (since no one is
there to teach you after your graduation), and quickly catch up with the new
technologies; otherwise, you may not be able to survive in the technical
- Thinking capability
- Independent thinking capability
- Independent thinking means that one should have his/her own views.
- Critical thinking capability
- Critical thinking means that
one should relentlessly attack the existing work/theory from all possible angles.
Attack its limitations and weaknesses or even tend to disprove the
theory. Unless the theory is completely justified, do not accept the
theory. After identifying the weaknesses of the theory, he/she is
ready to think creatively.
- Creative thinking capability
- Creative thinking means that one makes efforts to produce
something which does not exist in the literature (e.g., insights, simpler
proofs, new theorems).
- Logical reasoning
- Deductive reasoning:
- reasoning from the general to the special;
- can be trained by learning how to prove mathematical results (theorems,
- Inductive reasoning:
- reasoning from the special to the general;
- need to be able to identify the key common patterns in the observed
- there could be infinite number of possible patterns;
- how to identify meaningful patterns is hard;
- this is similar to knowledge discovery from a huge amount of data (a.k.a.
- Abductive reasoning or plausible reasoning:
- Plausible reasoning is a capability of identifying the hidden factors that
cause the observed phenomena.
- Plausible reasoning is crucial in knowledge discovery.
- Capability of doing independent and innovative research
- Writing skill, including writing journal/conference papers and proposals
- Communication skill (able to make your ideas clear to others)
- Oral presentation skill (able to give a good talk and attract people's
- Networking skill (good at socializing with professionals, government
agencies and industry sponsors)
- skill of managing a team
- being a visionary
- Types of thinking
- Forward thinking (deductive)
- Reverse thinking (inductive)
- Given a problem, come up with various methods
to solve it (need brainstorming)
- Analogy: Given the peak of a mountain, try to
find a path from the foot of the mountain to reach the peak.
- Think about whether the problem is ill-posed;
if so, reformulate the problem.
- Vertical thinking:
- Lateral thinking: multidisciplinary, cross
|Looking for the right approach
||Looking for as many approaches as possible
|Proceeds if there is a direction
||Proceeds to generate direction
||Is provocative (brainstorming)
||Can make jumps
|One must be correct at every step
||One does not have to be correct at every step
|Uses negative to block off certain pathways
||There is no negative
|Excludes what is irrelevant
||Welcomes chance intrusions
||Labels may change
|Explores most likely paths
||Explores least likely paths
|Is a finite process
||Is a probabilistic process
and Creative Thinking
- George Polya, ``Mathematical discovery: on understanding, learning, and teaching problem solving,'' New York: Wiley,
- George Polya, ``How to solve it : a new aspect of mathematical method,''
2d ed., Princeton, N.J. : Princeton University Press, 1973.
- George Polya, ``Mathematics and plausible
reasoning,'' Princeton, N.J.: Princeton University Press, 1954.
Associative learning and thinking, Jack Deal.
- Diane F. Halpern, "Thought and Knowledge: An
Introduction to Critical Thinking," Lawrence Erlbaum Associates; 4th edition
(July, 2002), ISBN: 0805839666.
Writing and Publishing
Writing on Henning Schulzrinne's web.
Hints and common
mistakes on Henning Schulzrinne's web.
- Criteria for Judging IEEE Prize Paper Awards,
by IEEE Paper Awards Committee
How to Organize your Thesis, by John W. Chinneck.
Advice to Authors of Extended Abstracts, by William Pugh.
on good mathematical writing, by David Goss
A primer on mathematical writing, by Steven L. Kleiman
How To Have Your Abstract Rejected,
by van Leunen and Lipton.
An Evaluation of the Ninth SOSP Submissions, or, How (and How Not) to
Write a Good Systems Paper by Roy Levin and David D. Redell
How to Get a Paper Accepted at OOPSLA, by Alan Snyder.
Advice for 1996 POPL submissions
- How to write
a white paper
- 4 levels of excellence in writing and the corresponding
- Level 1: Grammar
- Features: There should be no grammatical errors.
- Training: Develop grammar-awareness through intensive practice.
- Level 2: Coherence and logic flow
- Features: Logic flow should be smooth. There should be no jump in the
- Training: Learn how to create flow between sentences, how to make
transitions between paragraphs, how to present your thinking process.
Start from things known to things unknown. The lower level where the
writing starts, the better. Begin at the beginning. Don't start from
the middle, with an assumption that the reader knows the beginning. A common
fault of a writer is to assume that the reader has much knowledge about the
subject to be addressed.
- Keywords: coherence, cohesion, connection, transition, flow, evolution of
- Level 3: Rhetoric and diversity in vocabulary and expressions
- Features: Grace, elegance, giving an enjoyable reading.
- Training: Build vocabulary and learn rhetoric through reading articles that
- Level 4: Achieving the objective (marketing, convincing the reader, selling
your ideas, etc)
- Features: conciseness, giving compelling reasons, persuasiveness, and
- Training: Develop skills to make a persuasive, concise, logical and clear
presentation. The paper is expected to make compelling reading.
- Organization of a grant proposal (writing a grant proposal is
quite different from writing a technical paper)
- What problem?
- What approach?
- Advantages of the approach
- Potential impact of the proposed work
- Be concise; can cite your own paper for further details. Reviewers
do not have much time to read detailed descriptions in the proposal.
- Apply the old approach (developed in your thesis) to new problems, which
are significant and interesting.
- Related work
- J. M. Williams, "Style: ten lessons in clarity and grace," 7th ed., Longman,
- T. N. Huckin and L. A. Olsen, "Technical writing and professional
communication for nonnative speakers of English," 2nd ed., McGraw-Hill, 1991.
- Effective communication: tips on technical writing [pdf]
IEEE Signal Processing Magazine, Volume: 25, Issue: 3, Page(s): 129-132,
Date: May 2008
- Some good habits
- Write a note in HTML after meeting with the advisor
- Write a report in HTML or Latex every week so that you can see your
progress made. The report describes what the problems are and how
you approach the problems. The report will become a
journal/conference paper later.
- Write down the problems you have encountered. Use the techniques
in the book of "How to solve it" to deal with it.
- Simplifying the problem (solve the simpler problem first, then solve
- Divide and conquer
- Find lower/upper bounds
- Removing dependence between events
- How can you judge your simulation results are correct, instead of results
of buggy programs?
- Use rough analysis to get approximation or lower/upper bound and check your
simulation results are feasible.
- For example, a linear program may not have analytical (close-form) solution.
However, when you remove some constraints, the new linear program may admit a
close-form solution. The objective value corresponding to this
solution can be used as a lower or upper bound for the objective function of the
original linear program.
- Use numerical analysis to get rough results to compare with the simulation
- Simplify the constraints so that you can can rough analytical results to
compare with the simulation results.
- On the other hand, analytical results also need validation by simulations.
- Minimize the critical path by appropriately allocating resources among
nodes. Sometimes, you solve one problem but create another problem.
What is the right tradeoff? A smart design is to shift the burden in
a right way, e.g., shifting the burden from memory to processor so as to
minimize the delay.
Graduate Study in the Computer and Mathematical Sciences: A Survival
Manual, by Dianne O'Leary
How to be a Good Graduate Student/Advisor, by Marie desJardins
A Letter to Research Students, by Duane A. Bailey
How to do Research in the MIT AI Lab, ed. David Chapman
- The IUCS
Graduate Student Survival Guide.
- "A Stroke of Genius: Striving for Greatness in All You Do" by Richard
A good talk is the one that can make the ideas presented crystal clear to the
- If you lose your audience in the first two minutes, you will lose them in
the whole talk. So the first two minutes is very important and
is called honey-moon period. You should try every effort to
attract their attention. Use simple language (and figures) to
- what the problem is,
- what the motivation of your research is, and
- why it is an interesting research issue.
- The structure of your talk may be completely different from your
paper. In your paper,
- you can assume every reader is an expert in your area;
- you can assume that a reader can go back and forth through your paper
(like having a large brain memory);
- you may make claims by referring to other publications;
- you can use many techniques (e.g., use footnote and refer to later
sections or appendix) to improve a reader's
- you can write 30 pages and cite 50 references to present your work.
But in a talk,
- you need to adapt the level of the content to your targeted
- you should assume they can only remember a few previous slides up to
the slide that you are currently presenting (since they may not be
familiar with your area), so you may need to repeat some contents when
- you need to provide concrete and convincing evidences in the
talk since people may not accept/understand your claims (e.g.,
"other methods are complex but my method is simple" needs
convincing justification). So lots of effort is needed to
find/construct such concrete and convincing evidences to justify your
claims, when you prepare your slides.
- you may not be able to put all related material together in one slide;
so you may need to use words like "defined later",
"described later" or "explained later" when new
concepts appear in the slide but there is no space to describe
- you only have a limited time (say, 30 minutes) to present your work;
you need to make sure all important information about your work appears
in the slides; the content should not be superficial (since you need to
show depth to impress the audience) and cannot be too thorough
(otherwise you don't have enough time to present
them). You need to try every effort to strike a balance
between these two competing objectives.
- You talk is like telling a story; a story about how you explored the
problem. You need to use plausible/inductive reasoning, rather than
deductive reasoning (directly present the final results). The
presentation flow should be natural and logical to the audience
who is not familiar with your work. Presenting your thinking
behind your work or your design philosophy would be very helpful for the
audience to understand your work.
- Do not jump in your presentation. Make the presentation-flow smooth. Every sentence/figure appeared in your slides should be
followed easily by your targeted audience. Then after the talk,
your work will be crystal clear to the audience.
- Every claim needs justification or reference.
- If you can provide intuition while presenting theory
precisely, it is a good talk.
- Whenever presenting a plot, do not forget to describe what the x-axis and
A Top-Down Approach to Presenting Mathematical Theorems
- Present the basic idea and sketch the proof. (Motivate the audience.)
- Present the formal idea and the complete/rigorous proof. If
the proof is too complicated, only show an incomplete proof and leave the
rigorous proof as a self-study reading.
- Look back. Provide the intuition about the
theorem. Indicate the possible impact/applications of the
Common problems that an unseasoned speaker may have:
- Some assumptions are not given.
- Some notations are not well explained. Too many notations could confuse the audience.
- The level of the content is not accessible to those who are not familiar with your area. (The speaker may need to adapt the
level/emphasis during the talk.)
- The slides do not follow logically.
- There are thought-jumps in the presentation.
- Lack examples. Or the examples are not simple or not easily understandable.
- Key points are not emphasized adequately. The time spent on each slide should be somehow proportional to its importance in the whole
- Do not know how to handle questions appropriately. People may ask
unexpected questions. For example, people may ask you what if you
change your assumption; how to apply your approach to another (unexpected)
situation; how to solve a new problem. The questions may get out of the
focus of your talk and an unseasoned speaker may get lost in the questions.
- Be polite but stick to the focus of your talk.
- If you have a ready solution, tell the questioner. If not, you
could point out possible directions and then get back to your topic.
- To answer "what if you change your assumptions", you may say: "changing
the assumptions may result in different results and it may require
intensive study. I do not have a ready answer."
- Create suspense to raise the curiosity of the audience.
- Use stress and intonation to emphasize the key points.
- Produce captivating presentation by visual aids
(coloring, animation, videos, images, figures, etc.), which help guide the
attention of viewers and effectively get across your message.
- Use jokes/anecdotes to relax the audience and make your talk an enjoyable experience.
of an excellent instructor
- Networking on the Network by Phil Agre
- Thomas Hull, Michael A. Jones, and Diana M. Thomas, Interviewing for a job
in academia, Notices of the American Mathematical Society, November 1998,
pages 1353-1357. Emphasizes the questions you should have ready answers to.
- Mary Corbin Sies,
Academic job interview advice. A professor's advice based on experience.
Plus more questions:
- Adviser, Teacher, Role Model, Friend: On Being a Mentor to Students in
Science and Engineering, National Academy Press, 1997. So you'll know what
good advising is.
- Computer Science Faculty and Research Positions
The Young Scientists' Network
CRA Committee on the Status of Women in Research
SIGMOD academic careers information
Job hunting advice for scientists
Related Topics and Resources
Information resources for graduate students by Jennifer Myers.
A Guide for New Referees in Theoretical Computer Science,
by Ian Parberry.
On Being A Scientist: Responsible Conduct In Research,
from the National Academy of Sciences
- Papers on
women in computer science.
Study, Research, and Writing Skills web page
from the American Communication Association.
Information for current and prospective graduate students by
A Guide for Applying to Graduate Schools
by Piroz Mohseni.
- Ivan Sutherland, "Technology and Courage," in CMU Computer
Science: A 25th Anniversary Commemorative, ed. Richard F. Rashid. ACM
- Alan Jay Smith, "The task of the referee," IEEE Computer, April 1990,
- Free software for creating tables in Latex: