Statistical Approaches to Learning and Discovery

[ home | schedule | texts | assignments | readings ]

---

The course project is intended to give students a chance to exercise some of the ideas and methods discussed in the course on more realistic and challenging problems/data than considered in the problem sets. Define a narrow enough statistical problem that you can make significant progress on it in a three-week period. It's easy to go both overboard and underboard here -- a rough guideline is that the work involved should be about the equivalent of 2-3 problem sets, but that the project should involve "real models and real data." The projects can be done either individually or in groups of two.

A writeup on the project is due in class on May 8. You should hand in a research-style paper, approximately 5-10 pages in length. This writeup should include:

• A concise problem statement
• A description of related work
• A detailed presentation of the statistical techniques that you use
• Experimental results
• Conclusions
• References

• UC Irvine Machine Learning repository, and links to other resources.
• DELVE site for evaluation of learning experiments, University of Toronto.
• Case studies in data analysis from the Statistical Society of Canada.
• Mixture model resources, McMaster University.