• Compressed sensing -- implementations and applications
• Finding a hidden partition in a graph -- extending some existing theorems
• Modern regularized boosting algorithms -- implement and compare
• Maximum entropy discrimination -- implement for an application, or study a more complex prior
• Exponential family models over permutations -- the model selection problem (implementation or theoretical study).
• MAXIMIZATION of convex functions

You must write a description of your project: problem, how you approached the solution, results. Make sure that the description is clear and detailed enough for me to understand and grade. As a recommended size, the project should take 1-2 pages for problem description, and some 4-5 pages for approach, algorithms, experiments, not including figures. In addition, make clear figures showing your results and demonstrating how your algorithms worked. Besides the "performance" at whatever task you choose, you should display information that illustrates other properties of the algorithm (e.g verifies it's correct). These will be taken into account in the grading.

2. On "exam day", Wednesday March 19th, I will meet for 20 minutes with each of you. During this I will ask questions about the project. You don't need to prepare a presentation, in fact you are not require to prepare anything else. But if you feel the need to bring in extra results, a computer to demonstrate a point, and so on, do so.

The available meeting times are as follows:

10:40 E. Rombokas
11:00 J. Gruhl
11:20 Hoifung Poon
11:40 E. Herbst

4:30
4:50
5:10
5:30Bhushan Mandhani
5:50
6:10

Send me mail with the time slot you prefer. (There will be 8 projects total).