Location: MIC-Arena, Maria-Goeppert-Str. 3, 23562 Lübeck
Title: Large-scale numerical optimization with applications in computed tomography
Abstract: Numerical optimization plays an important role in reconstruction methods for computed tomography. While many problem formulations lead to optimization problems that are conceptually simple to solve, the large-scale nature of tomography problems makes them challenging and calls for computationally cheap methods. In this talk, we discuss two classes of such methods, namely the class of incremental proximal gradient methods and block coordinate descent methods. These methods can be viewed as generalizations of some classical tomography methods and can be applied to a wide range of problem formulations. We illustrate the properties of these methods based on some examples from tomography.