1: \begin{abstract}

2: We introduce new optimized first-order methods

3: for smooth unconstrained convex minimization.

4: Drori and Teboulle~\cite{drori:14:pof} recently described a numerical method 

5: for computing the $N$-iteration optimal step coefficients 

6: in a class of first-order algorithms

7: that includes gradient methods,

8: heavy-ball methods~\cite{polyak:64:smo},

9: and Nesterov's fast gradient methods

10: \cite{nesterov:83:amo,nesterov:05:smo}.

11: However, the numerical method in~\cite{drori:14:pof}

12: is computationally expensive for large $N$,

13: and the corresponding numerically optimized 

14: first-order algorithm in~\cite{drori:14:pof}

15: requires impractical memory and computation

16: for large-scale optimization problems.

17: In this paper,

18: we propose optimized first-order algorithms

19: that achieve a convergence bound

20: that is two times smaller than for Nesterov's fast gradient methods;

21: our bound is found analytically and refines the numerical bound in~\cite{drori:14:pof}.

22: Furthermore, the proposed optimized first-order methods

23: have efficient forms

24: that are remarkably similar to Nesterov's fast gradient methods.

25: 

26: %Include keywords, PACS and mathematical

27: %subject classification numbers as needed.

28: \keywords{First-order algorithms \and Convergence bound 

29: \and Smooth convex minimization \and Fast gradient methods}

30: % \PACS{PACS code1 \and PACS code2 \and more}

31: % \subclass{MSC code1 \and MSC code2 \and more}

32: \end{abstract}