There are two different homework options for the course:

  1. Do all 3 Individual Classifier Homeworks and the Ensemble Learning Homework, as described below.
  2. Learning Game Player
    1. Build the Learning, Game-Playing system and
    2. Do ONE of the Individual Classifier Homeworks. However, the classifier that you implement for part 2 CANNOT be the same learning mechanism that you use for part 1. So if you use a decision tree classifier for part 1, then you'll need to implement a Bayesian or Neural Network for part 2.

The Individual Classifier Homeworks

  1. Bayesian Network
  2. Decision Tree
  3. Neural Network

Combining the Classifiers into an Ensemble Learning System


A Learning, Game-Playing System

  1. Build an AI system to either play a k-person game (k > 1), such as Poker, Checkers, Othello, etc., or to solve (1-person) puzzles of a certain type (e.g. Sudoku, Cryptoarithmetic, etc.). Demonstrate that the system can solve many puzzles or play many games (against human and computer opponents). The system should be reasonably competent (i.e. it should only make legal moves and should not make too many ridiculous moves), though it need not be superior.
  2. Add in a machine-learning component to your AI system and demonstrate that, by using this module, the system can improve its performance over time. The learning module can be one of those described in the Classifier Homeworks, or it can be another of your choosing, such as reinforcement learning or genetic algorithms.

Last modified: Tue Jan 17 11:23:08 MET 2006