An affine scaling projective reduced Hessian algorithm for minimum optimization with nonlinear equality and linear inequality constraints (Q2485615)

The author proposes a modification of the affine scaling projective reduced Hessian trust region algorithm for solving optimization problems of the following form: \[ \text{Minimize }f(x)\text{ subject to }c(x)= 0,\;Qx\geq b, \] where \(f: \mathbb{R}^n\to\mathbb{R}\) is a smooth nonlinear function, \(c: \mathbb{R}^n\to\mathbb{R}^m\) is a nonlinear vector-function, \(Q\) is an \((n,k)\)-matrix and \(b\in\mathbb{R}^k\). The modification consists in the usage of a non-monotonic interior point backtracking strategy. Conditions under which the global convergence as well as fast local convergence of the suggested modification can be proved are formulated and the corresponding convergence of the algorithm is proved. Moreover, the author suggests a procedure, which overcomes the so-called Maratos effect and speeds up the convergence in some ill-conditioned cases.