- Supervised and semisupervised learning (1-2 lectures)
- an overview
- Classification: generative vs. discriminative models
- Basic theorems
- Where does optimization fit in?
- Unconstrained optimization (3 lectures)
- Boosting as gradient descent (2 lectures)
- Convex sets and functions. Examples from statistics. (4 lectures)
- Entropy and information.
- exponential family models, the maximum entropy principle, Bregman divergences (applications to clustering)
- [time permitting] principles of approximate inference in graphical models
- Convex constrained optimization problems and duality (3 lectures)
- Support vector machines as convex optimization problems (3 lectures)
- Compressed sensing, the LASSO and l1 regularization (2-3 lectures)
- Algorithms for constrained optimization (1-2 lectures)
- [time permitting] Conic programming, semidefinite programming and applications to modern kernel learning algorithms