25.2. Gradient Projection--Conjugate Gradient Method

The GPCG [(ref more-toraldo)] algorithm is much like the TRON algorithm, discussed in Section , except that it assumes that the objective function is quadratic and convex. Therefore, it evaluates the function, gradient, and Hessian only once. Since the objective function is quadratic, the algorithm does not use a trust region. All of the options that apply to TRON, except for trust region options, also apply to GPCG.