1: \begin{abstract} 

2: % Fast Iterative Hard Threshoding (FIHT) is a non-convex optimization algorithm widely used in compress sensing.

3: We propose a VHL-FIHT algorithm for matrix recovery in blind super-resolution problem in a non-convex schema. Under certain assumptions, we show that VHL-FIHT converges to the ground truth with linear convergence rate. %provided the number of samples is $\calO(s^2r^2\log^2(sn))$.

4: Numerical experiments are conducted to validate the linear convergence and effectiveness of the proposed approach.

5: \end{abstract}

6: