Learned Extragradient ISTA with Interpretable Residual Structures for Sparse Coding

Recently, the study on learned iterative shrinkage thresholding algorithm (LISTA) has attracted increasing attentions. A large number of experiments as well some theories have proved high efficiency LISTA for solving sparse coding problems. However, existing methods are all serial connection. To address this issue, we propose a novel extragradient based (ELISTA), which residual structure and theoretical guarantees. Moreover, most use soft function, been found to cause estimation bias. Therefore, function ELISTA instead thresholding. From perspective, prove that our method attains linear convergence. Through ablation experiments, improvements network verified in practice. Extensive empirical results verify advantages method.