1: \begin{abstract}

2: This paper considers the noisy sparse phase retrieval problem: recovering a sparse signal $\vct{x} \in \mathbb{R}^p$ from noisy quadratic measurements $y_j = (\vct{a}_j' \vct{x} )^2 + \epsilon_j$, $j=1, \ldots, m$, with independent sub-exponential noise $\epsilon_j$. The goals are to understand the effect of  the sparsity of $\vct{x}$ on the estimation precision and to construct a computationally feasible estimator to achieve the optimal rates. Inspired by the \emph{Wirtinger Flow} \cite{Candes14} proposed for noiseless and non-sparse phase retrieval, a novel thresholded gradient descent algorithm is proposed and it is shown to adaptively achieve the minimax optimal rates of convergence over a wide range of sparsity levels when the $\vct{a}_j$'s are independent standard Gaussian random vectors, provided that the sample size is sufficiently large compared to the sparsity of $\vct{x}$.

3: \\

4: \\

5: \textbf{Keywords}: High-dimensional $M$-estimation; Iterative thresholding; Minimax rate; Non-convex empirical risk; Phase retrieval; Sparse recovery; Thresholded gradient method.

6: \end{abstract}

7: