The effect of the step-size on the numerical behavior of a primal-dual interior-point algorithm applied to P*(κ) linear complementary problem

Abstract

Theoretically, the step-size plays a crucial role in the complexity analysis of primal-dual interior-point algorithms. In this paper, we would like to focus on the strategy of how to select the step-size. We propose three choices applied to P*(κ)-Linear complementarity problem based on two new kernel functions. The numerical behavior of the primal-dual interior-point algorithm is shown to be improved with these step-size choices. We have signicantly reduced the number of the inner iterations and the calculation time of the large-update algorithm.