The Elicited Progressive Decoupling Algorithm: A Note on the Rate of Convergence and a Preliminary Numerical Experiment on the Choice of Parameters

Abstract

The paper studies the progressive decoupling algorithm (PDA) of Rockafellar and focuses on the elicited version of the algorithm. Based on a generalized Yosida-regularization of Spingarn’s partial inverse of an elicitable operator, it is shown that the elicited progressive decoupling algorithm (EPDA), in a particular nonmonotone setting, linearly converges at a rate that could be viewed as the rate of a rescaled PDA, which may provide certain guidance to the selection of the parameters in computational practice. A preliminary numerical experiment shows that the choice of the elicitation constant has an impact on the efficiency of the EPDA. It is also observed that the influence of the elicitation constant is generally weaker than the proximal constant in the algorithm.