A numerical study of multiplier methods for constrained parameter optimization

Several algorithms based on the multiplier or augmented penalty function method for minimizing a function subject to equality constraints are surveyed and compared numerically. These methods employ an approximation of the Lagrange multiplier to increase efficiency. A revised method for choosing the multiplier is presented and compared numerically with the other algorithms. One scheme for automatically choosing the penalty constant associated with the algorithms is also presented.