|
1. | Unconstrained optimization
|
|
2. | Classification: bias, variance and applications of unconstrained optimization to classification. Bagging and Boosting. Boosting as gradient descent. [Note: Boosting was extensively discussed in STAT 591/EE 596 in Fall 06. This course offering will take care to minimize the overlap with the previous course.]
|
|
3. | Convex sets and functions. Examples from statistics: Entropy and information.
|
|
4. | More applications of convexity in statistics: principles of approximate inference in graphical models, the maximum entropy principle
|
|
5. | Convex constrained optimization
|
|
6. | Support vector machines as convex optimization problems
|
|
7. | [time permitting] Conic programming, semidefinite programming and applications to modern kernel learning algorithms
|