CompSci 175: Project in AI - Winter 2012

Instructor: Max Welling


Projects:

This year the default project is building a Mancala Game Agent
Mancala is one of the oldest known games.
The project will have elements of search (minimax search and alpha-beta pruning) and learning (e.g. reinforcement learning, neural networks)

Lectures on relevant topics will be provided by the instructor, such as constraint satisfaction, search and/or reinforcement learning.

Every game simulator should run from a website and beat the instructor of this course hands down.
A nice interafce is required (a 10 year old should be able to play and have fun).

Alternative projects are allowed but the teams will have write a proposal first and obtain permission from the instructor.
We have a few suggestions that we will discuss in class.


Teams:

Students are required to work in groups of 3 to 5 students (flexible on request).


Reports:

You are required to hand in a project report by March XX midnight, in the designated EEE dropbox.
Your report should explain the work you have done in your group. It is important that you
are very precise with respect to your contributions relative to the other groupmembers' contributions
Try to keep the report within 5 pages.

Reports are due March XX midnight. Please submit in pdf only.


Grading:

Class attendance is manadatory. Every class, teams will report on their progress and we will jointly discuss and brainstorm on how to proceed. Some classes will consist of a lecture.

Grading will also depend on what will be accomplished as a team. However, the instructor has the liberty to adjust grading on an individual basis.

If you choose this class, come motivated because it will only work if everyone will invest the necessary effort right from the beginning.


Slides:

A*-search [ppt] [pdf]
Minimax search with alpha-beta pruning [ppt] [pdf]

Reinforcement Learning [ppt] [pdf]


Software:

Code to Run Mancala Competitions

Dan Frost's Ucigame - Java Game Library