Next: Programming Projects
Up: E6998-002 High Performance Software
Previous: TA Page
There are three milestones for the research paper review:
- Choosing a research paper. You should find a paper in a recent
(2002 or later) computer science conference or journal, and get the
instructor's approval for that paper. The paper must:
- Be a full paper (e.g., at least 10 pages for a conference paper)
and not a short paper or poster paper.
- Be related to the course.
- Not be one of the papers the instructor will be using for course
material.
- Be on a topic different from your Survey Paper (see below).
You must finalize your choice and get the instructor's approval by
2/6/2007. You can get the instructor's approval either
before/after class, by email (send a link to the paper), or during
office hours.
- Writing a one-page summary of the paper. You should use 10-point
font and standard margins. Two-column format is acceptable. In this
summary you should describe:
- What problem is this paper trying to address? Why is the problem
important?
- What are the key ideas proposed in the paper, and why are they novel?
- How are the ideas evaluated? Is the evaluation fair/appropriate?
- Do you think the paper will have high impact? Did you like the
paper?
The one-page summary is due at the start of class on 3/20/2007,
in hardcopy.
- Giving a presentation to the class based on the paper.
You should prepare slides for the presentation. (In some cases
authors put presentation slides on the web. I would prefer that you
not use such slides, except perhaps to cut and paste performance graphs etc.
I'd like to see each of you have the experience of
preparing your own presentation.)
The length of the presentation will be determined later: It will
depend on the number of students taking the class.
Presentations will be in class on April 3 and 10.
Next: Programming Projects
Up: E6998-002 High Performance Software
Previous: TA Page
Ken Ross
2007-04-23