1: \begin{abstract}		

2: 		We present a novel way of generating Lyapunov functions for proving linear convergence rates of first-order optimization methods. Our approach provably obtains the \emph{fastest} linear convergence rate that can be verified by a quadratic Lyapunov function (with given states), and only relies on solving a small-sized semidefinite program. 	

3: 		Our approach combines the advantages of performance estimation problems (PEP, due to \citet{drori2014}) and integral quadratic constraints (IQC, due to \citet{lessard2016}), and relies on convex interpolation (due to \citet{taylor2017smooth,taylor2017exact}).

4: 	\end{abstract}

5: