Data Mining and Machine Learning Case Study


Course Outline

Usually in machine learning or data mining courses we learn many classification, clustering, or data processing techniques. However, there is no standard procedure on applying them to practical applications. Very often a painful process of trial and error is needed. While handling practical data may be an art but not science, in this course, we try to gain experiences from some past and existing data competitions: NIPS 2003 feature selection challenge, TRECVID video concept detection, and Netflix movie recommendation competition. Due to the instructor's background, we focus more on the use of machine learning methods, but not designing features for various applications. We will run this course in an interactive way, so students must present their findings every week. We will accept no more than 15 students.

Course Format

In the beginning of each project, the instructor gives a two-hour lecture introducing the project and possible plans. We then spend the following hour on dispatching jobs. In subsequent weeks, students present their progress (15 minutes per student or group), and we adjust our working plan.

If some techniques are needed and are new to most of us, some of you may be asked to give short introductions.


Possible Projects


Exams

No exams

Grading

It will be based on your results and presentations every week.
Last modified: Fri Mar 9 06:06:32 CST 2007