A class of collinear scaling algorithms for bound-constrained optimization: Derivation and computational results (Q1026441)

Algorithms for the approximated solution of a bound-constrained minimization problem are described. The construction of a local conic model for the barrier function and various alternatives for defining the corresponding quadratic subproblems are discussed. Methods for solving the quadratic subproblems and convergence criteria are examined. Algorithm schemes are presented for the outer and inner iterations of a proposed conic model trust region barrier algorithm. Some numerical results are obtained by applying various versions of the proposed algorithm to a variety of test problems.