In this talk, we will discuss extending the optimal gradient methods for solving convex optimization to deal with more general nonlinear, possibly nonconvex and nonsmooth, optimization problems. These algorithms will treat the nonconvex and convex optimization problems in a unified way so that they will achieve the best known complexity for solving a more general class of optimization problems. Both the well-studied techniques for nonlinear programming and the techniques for convex optimization can be combined in the framework of the discussed methods to improve their numerical performance. Joint work with Saeed Ghadimi and Guanghui Lan