|
Abstract
|
We give a simplified analysis and an improved iteration bound of a full Nesterov–Todd
(NT) step infeasible interior-point method for solving symmetric optimization. This
method shares the features as, it (i) requires strictly feasible iterates on the central path
of a perturbed problem, (ii) uses the feasibility steps to find strictly feasible iterates for
the next perturbed problem, (iii) uses the centering steps to obtain a strictly feasible
iterate close enough to the central path of the new perturbed problem, and (iv) reduces
the size of the residual vectors with the same speed as the duality gap. Furthermore, the
complexity bound coincides with the currently best-known iteration bound for full NT
step infeasible interior-point methods
|