1: \begin{abstract}

2: \modk{In this manuscript we study the properties of a family

3: of a

4: second order differential equations with damping, its

5: discretizations

6: and their connections with accelerated optimization

7: algorithms for

8: $m$-strongly convex and $L$-smooth functions. In

9: particular, using

10: the Linear Matrix Inequality {\color{black}(LMI)} framework

11: developed by  \emph{Fazlyab et. al. $(2018)$},

12: we derive analytically a (discrete) Lyapunov function for a

13: two-parameter family of Nesterov optimization methods,

14: which allows

15: for \modkzz{the} complete characterization of their

16: convergence rate. {\color{black} In the appropriate limit,  this

17: family

18: of methods may be seen as a discretization of a family of

19: second

20: order ordinary differential equations for which we construct

21: (continuous) Lyapunov functions by means of the LMI

22: framework. The continuous Lyapunov functions may

23: alternatively be obtained by studying the limiting behaviour

24: of their discrete counterparts.} % end of blue

25: Finally, we show  that the majority of typical discretizations

26: of the {\color{black} of the family of ODEs}, such as the Heavy

27: ball method,  do not possess {\color{black} Lyapunov functions with properties similar to those of the Lyapunov function constructed here for the Nesterov method.}}

28: 

29: \end{abstract}

30: