A polynomial-time inexact interior-point method for convex quadratic symmetric cone programming

Abstract. In this paper, we design an inexact primal-dual infeasible path-following algorithm for convex quadratic programming over symmetric cones. Our algorithm and its polynomial iteration complexity analysis give a unified treatment for a number of previous algorithms and their complexity analysis. In particular, our algorithm and analysis includes the one designed for linear semidefinite programming in "Math. Prog. 99 (2004), pp. 261-282". Under a mild condition on the inexactness of the search direction at each interior-point iteration, we show that the algorithm can find an ϵ-approximate solution in O(n 2 log(1/ϵ)) iterations, where n is the rank of the underlying Euclidean Jordan algebra