Machine Learning - Spring 2009

The main avenue of communication for this course will be through WebCT. Once you are registered, you should have access to the course in WebCT starting approximately one week prior to the start of class and should check in everyday for announcements. The main purpose of this page is to give some information for students considering taking the course. Because enrollment was at a manageable level, we dispensed with the overhead of WebCT and used email for communication.

Machine learning covers a vast array of algorithms, techniques and theory. Many of the topics merit their own courses. Because of this, we will be somewhat limited in what we can cover in the lectures. Lectures will be supplemented by homework assignments and additional reading. Potential algorithm topics include: decision trees, rule mining, nearest neighbor, feedforward neural networks, evolutionary computation, ant colony optimization, particle swarm optimization, reinforcement learning, bayesian networks, inductive logic programming, support vector machines, inductive logic programming and case based reasoning. Theory and techniques will be covered as needed.

The course will be taught from the perspective of a practicing engineer who may need to implement these algorithms and know their strengths and weaknesses. The course will give you some experience turning math in the practical programs. Consequently, there will be a lot of programming in this course. The preferred implementation language will be Java 5 or older (Java 6 users should target the Java 5 JVM). Good softare design and implementation will be expected.

At this time, it is anticipated that there will be a research ("term") paper, weekly assignments, and a "big picture" final exam.