1: \begin{abstract}

2:  In this overview article we will consider the deliberate restarting of

3:  algorithms, a meta technique, 

4:  in order to improve the algorithm's performance, e.g., convergence

5:  rates or approximation guarantees. One of the major advantages is that

6:  restarts are \emph{relatively} black box, not requiring any

7:  (significant) changes to the base algorithm that is

8:  restarted or the underlying argument, while leading to \emph{potentially significant}

9:  improvements, e.g., from sublinear to linear rates of

10:  convergence. Restarts are widely used in different fields and have

11:  become a powerful tool to leverage additional information that has

12:  not been directly incorporated in the base algorithm or argument. We

13:  will review restarts in various settings from continuous

14:  optimization, discrete optimization, and submodular function

15:  maximization where they have delivered impressive results. 

16: 

17: \keywords{restarts  \and convex optimization \and discrete

18:   optimization \and submodular optimization.}

19: \end{abstract}

20: