Numerical experiments with a continuous L-2-exponential merit function for semi-infinite programming

Here, we present some numerical experiments with a reduction method for solving nonlinear semi-infinite programming (SIP) problems. The method relies on a line search technique to ensure a sufficient decrease of a L_2-exponential merit function. The proposed merit function is continuous for SIP and improves the algorithm efficiency when compared with other previously tested merit functions. A comparison with other reduction methods is also included.The authors are grateful to three anonymous referees for their comments and suggestions. This work has been partially suported by the Algoritmi Research Center and by Portuguese FCT grant POCl/MAT/58957/2004