An accelerated forward-backward splitting algorithm for solving inclusion problems with applications to regression and link prediction problems

Abstract

The forward-backward method is a very popular approach to solve composite inclusion problems. In this paper, we propose a novel accelerated forward-backward algorithm to obtain the vanishing point of sum of two operators in which one is maximal monotone and other is M-cocoercive, where M is a linear bounded operator on underlying spaces. Our proposed algorithm is more general than previously known algorithms. We study the convergence behavior of proposed algorithm under mild assumptions in the framework of real Hilbert spaces. We employ our model to solve regression problems and link prediction problems for high dimensional datasets and conduct numerical experiments to support our results. This model improves convergence speed and accuracy in respective problems. We also conduct numerical experiments to support our results.