Validated infeasible interior-point predictor-corrector methods for linear programming

Many problems arising in practical applications lead to linear programming problems. Hence, they are fundamentally tractable. Recent interior-point methods can exploit problem structure to solve such problems very efficiently. Infeasible interior-point predictor-corrector methods using floating-point arithmetic sometimes compute an approximate solution with duality gap less than a given tolerance even when the problem may not have a solution. We present an efficient verification method for solving linear programming problems which computes a guaranteed enclosure of the optimal solution and which verifies the existence of the solution within the computed interval.