Python - Builtin Object Types

Download Report

Transcript Python - Builtin Object Types

Technical Writing
Vikram Pudi
Dedicated to:
My Ph.D advisor
Prof. Jayant Haritsa
IISc, Bangalore
2
Vikram © IIIT
Phases in Research
One Year
 Find area to work in
 Read papers
 Extract / formulate high impact problems
 Search for related work
 Usually related work is not good enough
3
Vikram © IIIT
The Interesting Part
One Year
 Invent your own solutions
 In CS research, this translates to



4
algorithm design / analysis
coding
night-outs and coffee
Vikram © IIIT
The Boring Part
3 years
 Write reports, papers, thesis, …
5
Vikram © IIIT
But it is important…



6
It is your duty as a scientist to share your
discoveries with others
Your work is understood only from what
you present / publish
You are evaluated only based on what you
write in your reports, etc.
Vikram © IIIT
It is actually exciting!


7
It is your work
You want others to feel the excitement you
felt when you invented the solutions
Vikram © IIIT
It has ingredients of programming

Modularity






8
As mentioned in Section 2, …
As mentioned in the Introduction, …
Minimize redundancy
Linear – don’t jump back and forth
Clarity, elegance and flow
Choose good titles (like variable names)
Vikram © IIIT
Popular Misconceptions



9
Should have lot of math
Should be hard to read
Should fill all 20 pages
Vikram © IIIT
Reality

Math should be used only where it is required



Reviewers don’t like reading hard-to-read papers



academic readers know that hard-to-read doesn’t mean weighty
matter
will be able to detect when a writer is just showing off
Quality is more important than quantity



10
to make things more precise / unambiguous
where equivalent English is too verbose
It is usually hard to fit your content in 20 pages
Be concise
Use figures and examples to fill space if necessary
Vikram © IIIT
Format of Papers / Reports


Title: Less than about 11 words
Abstract: 200 – 250 words



Introduction





One para – background / history
One para – motivation
One para – your contributions
One para – organization
Formal Problem Statement – some other name

11
Write this at the end
Mention what you have achieved; mention key expt results
Mention assumptions of setting under which your solution
applies
Vikram © IIIT
Format (contd.)

Background – some other name





Your Solution – in one or more sections
Related Work




12
Describe background necessary to understand the rest of the
document
Optional – point to other documents
Use to fill space if required
cite and briefly discuss other related work
mention how it is different from your work
mention their limitations but be polite
people like to see their work here; so mention anything that
seems related if done by a program committee member!
Vikram © IIIT
Format (contd.)

Performance Model


Experiments







13
Use graphs and tables to show results
Refer the graphs / tables in the text: We see in Figure 1 that …
Discuss intuition behind why the graphs are the way they are
Should sound like: Due to these reasons the results are as
expected.
Conclusions


Describe your experimental setup and metrics you used to
evaluate your solution against previous work
Like abstract, but some more technical points like merits of datastructures used, etc. – assume reader has already read thru the
document
Can mention possible future work
Acknowledgements
References
Vikram © IIIT
Format of a Thesis




Almost like a paper
Sections become chapters
Sub-sections become sections
In addition, it has:




14
a table of contents
list of figures
list of tables
index
Vikram © IIIT
Checklist



Do a spell-check
Ensure figure / table / section numbers match those in
text
Look at start and end of each page






15
E.g. A section title should be at the end of some page
Use \noindent after figures / tables
Ensure figures / tables are placed where you want them
Give good names for your algorithms / solutions – some
nice expansions
Use those names to make your title
Take a printout and read thoughly
Vikram © IIIT
Other tips



Use LaTeX
Review lots of good / bad papers; avoid their mistakes
Be ambiguous in your initial abstract


Be picky about how your sentences sound & feel





Like changing a software’s design
Finish a draft of the paper at least a week before the
final deadline

16
Can you write the same thing in a simpler way?
Can you write the same thing in a concise way?
Are you moving back and forth between two concepts?
A different word may be a better fit to explain this?
Expect to change your organization a couple of times


Just mention what you achieve without revealing techniques
Otherwise your chance of acceptance reduces to 20 – 30%
Vikram © IIIT
Take Home
Writing a paper requires effort, time,
an eye for detail, is addictive, exciting
and rewarding.
17
Vikram © IIIT