1: \begin{abstract}

2:    Recently, the study on learned iterative shrinkage thresholding algorithm

3:    (LISTA) has attracted increasing attentions. A large number of experiments as

4:    well as some theories have proved the high efficiency of LISTA for solving

5:    sparse coding problems. However, existing LISTA methods are all serial

6:    connection. To address this issue, we propose a novel extragradient based

7:    LISTA

8:    (ELISTA), which has a  residual structure and theoretical guarantees. In  particular, our algorithm can also provide the interpretability for Res-Net to

9:    a certain extent.

10:    From a theoretical perspective, we prove

11:    that our method attains linear convergence. In practice, extensive empirical results verify the advantages of our method.

12: \end{abstract}

13: