An improved iterative soft thresholding algorithm and application.

In order to solve the problems of the iterative soft thresholding algorithm (ISTA) that convergence rate is slow and the obtained optimal solution is not enough sparse, an improved algorithm called SFISTA based on the gradient algorithm is proposed. The algorithm modifies the gradient operator in the ISTA iterative method, so that the solution of the iterative point xⁿ⁺¹ depends on the iteration of the first two steps. Applying SFISTA to sparse signal processing and sparse principal component analysis, the experimental results show that the new algorithm not only can improve the convergence rate of ISTA, but can also promote the sparsity of the optimal solution. [ABSTRACT FROM AUTHOR]