1: \begin{abstract}

2: %\blue{Position CS as field with major impact.}

3: %Compressed sensing is a data acquisition theory for accurate signal reconstruction with significantly less number of samples than traditionally believed necessary when the signals are sparse. \blue{Position IRLS as key method for CS} In history, Iteratively re-weighted least squares (IRLS) iteratively solves a weighted least squares approximation problem in order to solve an unconstrained $\ell_p$ $(1<p\leq 1)$ problem. 

4: %	\blue{Identify theoretical gap, with words rather than math notation if possible.}

5: %	In  IRLS has been shown to be able to solve compressed sensing problem: $\min_{\Phi(x) = y} \norm{x}_1$. It is shown in [Ingrid Daubechies, Ronald DeVore, Massimo Fornasier, and C Sinan Gunturk. \emph{Iteratively reweighted least squares minimization for sparse recovery}. Communications on Pure and Applied Mathematics, 63(1):1–38, 2010] that, if the measurement matrix $\Phi$ has Null Space Property (NSP) of order $K$ for $0<\gamma<1$, any solution with sparsity $k$, can be recovered via IRLS if $k$ is smaller than $K$ with a positive gap, moreover, local linear convergence rate can be achieved for sufficiently small $k$. 

6: %	\blue{State contribution (filling gap)}

7: %	We show solution can be recovered for all $k \leq K$, and, in this case, local linear convergence can be achieved via a new version of the IRLS algorithm, where the $\epsilon$-updating rule is modified. \blue{An example that differentiates the methods is presented}

8: %\end{abstract}

9: