Orthogonal and conjugate basis methods for solving equality constrained minimization problems (Q1312090)

The authors describe and analyze two methods to construct the search direction during a step of the iterative process approximating solutions of equality-constrained minimization problems of the form \[ F(x) = \min!\mid c_ i(x) = 0,\quad i = 1,2,\dots,m. \] The very popular approach consists in calculating two orthogonal components of the search direction (i.e. tangential and normal to the constraints). The methods of this kind are called orthogonal basis methods. The second method, based on an idea given by L. C. W. Dixon (in a technical report of the Hatfield Polytechnics in 1987) and consisting in construction of the search direction from two moves which are conjugate with respect to the Hessian of the Lagrangian function, called conjugate basis method, is contrasted with the first one. The paper also contains an informal description of minimization algorithms for the two methods and the analysis of their local convergence properties, illustrated with some results of numerical experiments. The authors conclude that the (as theoretical as empirical) benefits of the second method encourage to the further investigations of the conjugate basis approach, and announce some new research tasks touched on by L. C. W. Dixon and the second author, related to the use of the so called restoration steps technique.