KIIS: Written assignment, part 2
While the first part of the written assignments consisted of fixed exercises to
train Prolog, the second part involves an individual choice of topic involving
some AI application.
You should agree on the choice of topic with your teacher before you start.
It is acceptable to work together 2-3 students on a topic, but the written answers
must be written and given in individually.
You should give in a written report of 4 to 7 pages at latest by
November 11, 2007. Send it as an email attachement as a
single file (preferably pdf, alternatively Word, and no separate program or data files,
please!).
If your text tends to be a bit longer that the 7 pages, don't spend lot of time to
abbreviate it, but you should be take care to limit the level of ambition so that
the assignment does not explode in time.
You are expected to give a short presentation (e.g. using powerpoint) for
the other students (and your teacher) at the last regular course day, November 13.
In case of related answers on a common topic, you can prepare a common presentation
in a way such that everyone contributes.
Here are some possible topics.
-
Write a small and illustrative application using the tools, we have used
(or will use) in the course, but for another example or application than
those we looked at already in examples and exercises.
For example:
- Abduction: Find another diagnosis or planning case and show its implementation
in Prolog+CHR.
- Use Saha's Neural Network tool for an interesting application;
notice that there are both a classification and a prediction version
(prediction; with numerical output).
- Natural language analysis. Define a grammar for a very simple subset of
a natural language with some specific sort of texts in mind. Define
some constraints which can produce a little knowledge base (as output
constraint store).
- Use PRISM to implement a Bayesian network for some classification problem.
Train the network with sample data (which you may have produced artificially
or found otherwise) and test the trained net.
- As above for a selected HMM.
- Explain some existing tool for one
of the tasks we have looked at in the course
(possibly more practical or more efficient than
the methods we have used). Test in on some very simple example - maybe chosing
from the course examples and exercises so you can make a comparison.
- There are several systems and packages for implementing
neural networks - select one, describe it in a few pages and test an example.
- Similarly for evolutionary computing
- Similarly for Bayesian networks or HMM, or...
- If you have an interest in, and preferably some background knowledge,
you are welcome to consider any other AI related topic
(but discuss with your teacher and agree on the subject beforehand),
Timing:
We should aim at that most student hare identified their topic and perhaps started
working by October 16, and all students by October 23.
Last modification 09-oct-2007,
Henning Christiansen