1: \begin{abstract}

2: %\textbf{NeuRIPS Abstract Submission Deadline: \emph{May 19}; Paper submission and co-author registration deadline: \emph{May 26}.}

3: %

4: Much recent interest has focused on the design of optimization algorithms

5: from the discretization of an associated optimization flow, i.e., a system

6: of differential equations (ODEs) whose trajectories solve an associated

7: optimization problem.

8: %

9: Such a design approach poses an important problem: how to find a principled

10: methodology to design and discretize appropriate ODEs.

11: %

12: This paper aims to provide a solution to this problem through the use of

13: contraction theory. 

14: %

15: We first introduce general mathematical results that explain how

16: contraction theory guarantees the stability of the implicit and explicit

17: Euler integration methods.

18: %

19: Then, we propose a novel system of ODEs, namely the

20: Accelerated-Contracting-Nesterov flow, and use contraction theory to

21: establish it is an optimization flow with exponential convergence rate,

22: from which the linear convergence rate of its associated optimization

23: algorithm is immediately established. Remarkably, a simple explicit Euler

24: discretization of this flow corresponds to the Nesterov acceleration

25: method.

26: %

27: Finally, we present how our approach leads to performance guarantees in the

28: design of optimization algorithms for time-varying optimization problems.

29: \end{abstract}

30: