1: \begin{abstract}

2:   Modern statistical applications often involve minimizing an  objective function  that may be nonsmooth and/or nonconvex.   This paper focuses on  a broad Bregman-surrogate   algorithm framework  including the   local linear approximation, mirror descent, iterative thresholding,  DC programming and many others

3:  as particular instances. The   recharacterization via generalized Bregman functions enables us to construct suitable error measures and  establish global convergence rates   for    nonconvex and nonsmooth  objectives in possibly high dimensions. For sparse learning problems with a composite objective, under some regularity conditions, the obtained estimators  as the surrogate's fixed points, though not  necessarily local minimizers,  enjoy provable statistical guarantees, and  the sequence of iterates can be shown to approach the statistical truth within the desired accuracy geometrically fast. The paper also studies how to  design adaptive momentum based accelerations without assuming convexity or smoothness by  carefully controlling stepsize and relaxation  parameters. %Simulations  are performed to support the theoretical results.

4: \end{abstract}