ITERATIVE METHOD FOR RANK-DEFICIENT (2,1)-BLOCK KKT SYSTEMS

Karush-Kuhn-Tucker (KKT) systems are widely used in scientific and engineering problems. They are characterized by a 2-by-2 block structure coefficient matrix that exhibits various features, such as sparsity, symmetry, non-symmetry, singularity, and nonsingularity. Solving singular (2,1)-block KKT systems presents significant computational challenges due to their inherent complexity. To address this challenge, an iterative method has been introduced explicitly for addressing singular (2,1)-block KKT systems. The proposed method reduces the system size by using only maximum linearly independent rows of (2,1) block matrices and effectively handles the singularity by transforming the row deficiency problem into a reduced form. The effectiveness and efficiency of the proposed iterative method are verified using numerical experiments with various matrices. It has demonstrated the capacity to provide definitive solutions for singular KKT systems and improve computing performance across multiple applications.