The Boosted DC Algorithm for Linearly Constrained DC Programming

Abstract The Boosted Difference of Convex functions Algorithm (BDCA) has been recently introduced to accelerate the performance classical (DCA). This acceleration is achieved thanks an extrapolation step from point computed by DCA via a line search procedure. In this work, we propose extension BDCA that can be applied difference convex programs with linear constraints, and prove every cluster sequence generated algorithm Karush–Kuhn–Tucker problem if feasible set Slater point. When objective function quadratic, any bounded R-linearly (geometrically) convergent. Finally, present some numerical experiments where compare on challenging problems: test copositivity given matrix, solve one-norm infinity-norm trust-region subproblems, piecewise quadratic problems box constraints. Our results demonstrate new outperforms DCA.