Optimering för maskininlärning, 7.5 hp
The following course program contains all relevant information about the course regarding content, logistics, deadlines etc.
The rest of this page will act as a repository for the course material and a place to make announcements when needed. This is the first time the course is given so documents will be updated with minor changes and additional material when needed.
The exam and hand-ins are now graded and the results will be registered in Ladok within the next couple of days. To view the exam, contact Martin M.
The exam with solution is now available. One clarification regarding problem 7 was made. Gamma was assumed to be positive, the problem is not solvable otherwise. No points deductions will be made for errors relating to this.
- L1 - Convex sets
- L2 - Convex functions
- L3 - Subdifferential and proximal operator
- L4 - Conjugate functions and duality
- L5 - Least squares
- L6 - Logistic regression
- L7 - Support vector machines and kernel methods
- L8 - Multiclass classification
- L9 - Deep learning
- L10 - Proximal gradient method
- L11 - Stochastic and coordinate proximal gradient method
- L12 - Scaling, Newton's method, quasi-Newton
- L13 - Coordinate and stochastic variations etc
- L14 - Recap
- Hand-In 1: manual, code
- Hand-In 2: manual, code, hints
- Hand-In 3: manual, code
- Extra Credit Hand-In: Choose one of the following, both are done individually.
Aids: Cheat Sheet