Add notes on paper presentations.

This commit is contained in:
Justin Hsu 2018-09-19 14:55:56 -05:00
parent f4c723c78b
commit 422261a922
4 changed files with 43 additions and 3 deletions

View File

@ -0,0 +1,39 @@
During this class, you will be responsible for presenting one paper in class and
leading the discussion. [Here](../resources/readings.md) is a list of suggested
papers; you are free to choose a paper that is not on the list, but please clear
it with me first. The day before each presentation, I will send out a few short
questions to check your understanding. Please reply with short responses---a few
sentences should be more than enough.
Research papers are written for a very specific technical audience. Papers
appearing in conferences---most papers in computer science---are also subject to
tight page limits and are typically extremely condensed; many things are left
unsaid. (Here is a useful
[guide](https://web.stanford.edu/class/ee384m/Handouts/HowtoReadPaper.pdf) to
reading papers.) When presenting a paper in class, you should not try to
compress the material from the paper. Instead, you should try to expand and
unpack the paper, so that it is easier to understand.
Here are a few specific things to keep in mind when presenting a paper.
- **Make sure the high-level picture is clear.** Make sure to explain the
problem the paper is trying to solve, the setting, and as much of the
motivation behind the paper as possible.
- **Don't spend the whole time presenting technical details.** It is probably
not interesting for the class to spend the whole presentation talking about
the technical details in a single proof.
- **You don't have to present the whole paper.** It is simply not possible to
present every detail in the span of one lecture. For some papers, it may not
even be possible to present each main contribution. Focus on the one or two
most important contributions (as decided by yourself). If the paper first
discusses a "core" or "basic" version, and then later adds on a bunch of
advanced extensions, *focus on the core version*.
- **Give as many examples as you can.** Most research papers are extremely
condensed, and do not have nearly enough examples. Present as many examples as
you can. Your examples don't need to be drawn from the paper---small examples
are the most useful, as simple as possible.
- **Keep the class background in mind.** Our class has students from a variety
of backgrounds; try to keep this in mind. It's better to briefly explain a
technical term if you are not sure everyone knows what it means. You should
assume that everyone in the class has basic familiarity with the paper (say,
assume everyone spent about 30-45 minutes reading the paper before class).

View File

@ -27,6 +27,7 @@ pages:
- Papers: 'resources/readings.md' - Papers: 'resources/readings.md'
- Software: 'resources/software.md' - Software: 'resources/software.md'
- Related Courses: 'resources/related.md' - Related Courses: 'resources/related.md'
- Project: - Assignments:
- Details: 'projects/details.md' - Presentations: 'assignments/presentations.md'
- Gallery: 'projects/final.md' - Project: 'assignments/project.md'
- Gallery: 'assignments/gallery.md'