This series of lectures will cover recent optimisation techniques for imaging. The specificities many imaging problems are

- their size (usually “large” but not “huge”),

- the structure of the data, on a 2 or 3D grid, which allows for very basic parallelism (as implemented in GPUs)

- their lack of smoothness

The lectures will therefore focus essentially on (mostly convex) optimisation methods for non-smooth problems: duality, proximity operators and proximal splitting, etc. We will discuss rates of convergence of first order methods (lower bound, “optimal” algorithms…) and try to describe practically useful algorithms and general convergence analysis techniques.