An adaptive augmented Lagrangian method for large-scale constrained optimization (Q494324)

An adaptive augmented Lagrangian algorithm is proposed for large-scale constrained optimization problems by updating the penalty parameters and/or Lagrange multipliers within the trust-region framework. The well-posedness and global convergence of the method are proved. The effectiveness of the adaptive penalty parameter updating strategy is illustrated in terms of the numbers of iterations and gradient evaluations on several test problems.