1: \begin{abstract}

2:   For  a functional $\E$ and a peak selection that picks up a global

3:   maximum of $\E$ on varying cones, we study the convergence up to

4:   a subsequence to a critical point of the sequence generated by a

5:   mountain pass type algorithm. Moreover,

6:   by carefully choosing stepsizes, we establish the convergence of the whole

7:   sequence under a ``localization'' assumption on the critical point.

8:   We illustrate our results with two problems: an indefinite

9:   Schrödinger equation

10:   and a superlinear Schrödinger system.

11: \end{abstract}

12: