CS888 Winter 2010

Project information

Adapted from Bill Cowan's project page from the Spring 2004 offering of CS888.

Tentative due date: April 21st, 2010. This date might be adjusted as we get a better idea of internal deadlines at the graduate office.

General Approaches

There are two obvious ways of doing a project.

1. Do a small implementation; present and/or write up the results.
2. Write a short research paper.

These are not the only acceptable project formats, only the two most obvious. It's a good idea to talk to us if you are thinking of a different idea. We will most likely be in favour of it, but you may be saving yourself from a costly blunder.

Implementation

Here is an example of a UI experiment that is about the correct level of complexity. It took me several hours to figure out what the experiment should do, and about six hours to implement and debug the Java applet. It takes less than five minutes for a subject to actually do the experiment, and about eight subjects is plenty. The times measured are long enough that we don't have to worry about running the experiment on a specially-configured computer. Data analysis, which I have done every time the experiment has been run as part of CS689, takes one or two hours.

The write up of experiments usually follows a standard format.

1. Introduction, which discusses the question that the experiment is supposed to answer.
2. Method, which describes the detailed procedure of the experiment, with short justifications explaining why unusual aspects of the procedure are done the way they are.
3. Results

Research Paper

Find a small (three or four) collection of papers describing similar experiments, possibly a series of experiments from a single research group, possibly competing papers from different groups. (The paper you presented would be an obvious starting point.) Read the papers and write a short comparison.

Scope and Scale

Time

We expect the project to take you ten to twenty hours, plus writing up whatever you hand in. Spending this time doing something unintelligent will, of course, not produce a good result. In addition, spending additional time doing more of the same unintelligent thing will not improve your mark. I suggest that you come and talk to us about how you are planning to spend your time before you start work.

Size

We expect you to hand in about 3,000 words, possibly less, not more. You will be marked based on the insight you show; anything that simply repeats or rewords stuff from lectures or papers is not likely to be mark-enhancing.

Marking

We normally mark on the following scale - owing to Stephen Mann.

Implementations

30 marks -

competent and almost complete, no new ideas

35 marks -

well done and complete

40 marks -

excellently done, paper extended with some new ideas

Research Papers

30 marks -

obvious papers described, no real synthesis

35 marks -

obvious papers well synthesized, obvious research proposed

40 marks -

more than obvious papers well synthesized, novel research proposed