Since the invention of turbo codes in 1993 there has been an enormous interest and progress in the field of capacity approaching code constructions. Many classical constructions have been replaced by newer, better performing codes with feasible decoding complexity. Most of these modern code constructions, such as turbo codes, Gallager's low-density parity-check (LDPC) codes and their generalizations, can be modeled by sparse graphical models. Spatial coupling of sparse graphical models has in the last years attracted a lot of interest due to the threshold saturation phenomenon, which leads to capacity achieving performance with iterative message passing decoding. Polar codes are a recently discovered class of capacity achieving codes that are formed by an explicit construction based on a phenomenon called channel polarization. These codes, too, have various low-complexity decoding algorithms based on message passing on a sparse graph that has a recursive structure similar to that of fast transforms in signal processing.