In this talk, we introduce a concrete algorithm for phase retrieval problem, which aims to recover a signal from phaseless measurements. In short, the algorithm, which we refer to as Gauss-Newton method, can be divided into two stages. In the first stage, the algorithm devotes to find a good initial estimation. The second stage of the algorithm is to update the iteration point by Gauss-Newton iteration. Here the initialization method can provide a good initial guess by using optimal number of measurements. For real-valued signals, we proved that a re-sampled version of the algorithm quadratically converges to the global optimal solution with the number of random measurements being nearly minimal.