The
quality of our discussions will rely on how prepared everyone is when they come
to class. It is important to do the
reading in order to actively participate.
Students are required to submit two paper reviews per week for the
assigned papers. (If we read more than 2
papers, just choose 2 to review.)
A
good review includes a brief synopsis of the paper content, a thoughtful
analysis of its contributions and strengths/weaknesses, and notes about points
that are unclear. Please address the
following points (in any order) in your review:
·
Give a brief
(2-3 sentences) summary of the paper in your own words.
·
What appears to
be the main contribution of the paper?
·
What are the
paper’s primary strengths? weaknesses?
(These may refer to both technical strategies and clarity.)
·
How convincing
are the experiments? If they appear
lacking in some way, what would you suggest be tested?
·
Are there any
obvious (or less obvious) ways this work could be extended?
·
Any additional
comments, including questions it raises for you, or points that are not clear.
Check
out Blackboard
> Course Documents > Review examples for examples of well-written
reviews.
Reviews
are due by 10 PM on the night before
class (Wednesday). Please post each
review using Google Docs (instructions are on Blackboard). In weeks that you are presenting or giving a
demo, it is not necessary to write any reviews.
Each
student will give a presentation in class covering 2-3 papers on a topic
selected from the course syllabus list.
This presentation should overview the papers and explain technical
details, but also synthesize any underlying commonalities or highlight interesting
distinctions. The goal is to become
familiar with some of the work in this particular area through the selected
papers, and to relay it to the class with a polished, well-organized talk.
The
presentation should be approximately 30 minutes and include these components:
·
clear statement
of the problem (and scope of background reading)
·
why the problem
is interesting, important, difficult
·
assumptions
applied
·
key technical
ideas, how they work, main contributions, strengths and weaknesses
·
means of evaluation:
How are the methods tested? What kind of test data?
·
open problems
and issues raised in the papers
·
specify
potential points of discussion for the class
The
key point is to synthesize the material and describe how the technical
contributions fit together when possible.
Use applications to motivate the work, and look for visual elements to
put in the presentation. Check out the
webpages linked on the class webpage, and also look at authors’ webpages for
supplementary materials. It’s ok to grab
a few slides from conference talks etc., but try to organize your presentation
around the themes/concepts, not just the separate papers.
For
each topic one person will present a “demo” or evaluation of some main idea in
a paper we read. When you are in charge
of the demo, basically your job is to implement a distilled version of an
essential technical idea in the paper, and show us some toy example of how this
works in practice. For a number of
papers, you may be able to find code or binaries provided by the authors
online. The goal is to help us gain a
more complete intuition about the work we are studying.
You
might:
·
experiment with
different types of training/testing data sets
·
examine the
method’s sensitivity to relevant parameter settings
·
show a
simplified example that highlights an expected strength/weakness of the
approach
Note
that the goal here is not to recreate published results or to build systems as
described in the paper. Instead, you are
looking to make a small illustrative demo that will let us more deeply
understand what we have read. Spend some
time playing with your implementation, and put thought into what would be an
instructive toy example to show the class.
The demo should allow us to learn something about the method, not just
see it. If you needed to implement
something yourself, explain how you did it, and especially point out any
details or choices that weren’t straightforward. Be sure to explain the rationale for the
outcomes, and conclude with a summary of the message(s) your example
illustrates.
A
demo presentation should take about 20-30 minutes. In addition to the presentation, make a
webpage to outline the demo and include links to any existing code, data, etc.
you may have used.
Timetable for presenters:
By
the Thursday the week before your presentation is scheduled:
-
Email slides to
the instructor, and schedule a time to meet and discuss.
The
week of your presentation:
-
Refine slides,
practice presentation, know about how long each part requires.
The
day of your presentation:
-
Send final
slides (and, for demos, pointer to webpage) to instructor.
As
part of this course, students will do research-oriented projects in pairs. A project could be built around any of the
following:
·
an extension to
one of the techniques studied in class
·
an in-depth
analysis and empirical evaluation of one or two related techniques
·
design of a
novel approach and accompanying experiments
Initial
project proposals will be due before the middle of the term. Proposal guidelines are here.