Asymmetric Bregman Forward-Backward Splitting with Projection Correction
This thesis examines first-order Bregman algorithms in a primal and a primal-dual setting. The Bregman gradient descent algorithm is introduced from a majorization-minimization perspective and as a generalization of the gradient descent algorithm. Concepts such as relative smoothness and Legendreness are defined and are shown to be natural restrictions in order to show convergence results. A speci
