A regularized variant of the continuous second-order gradient projection method (Q2563470)

The minimization problem \[ \inf J(u), u \in U, \] \[ U = \{ u \in U_{0}: g_{i}(u) \leq 0, \;i=1,\ldots,m;\;g_{i}(u)=0,\;i=m+1, \ldots, s \} \] is considered. Here \(U_{0}\) is a given convex subset of a Hilbert space \(H\) and the functions \(J(U), g_{1}(u), \ldots, g_{s}(u)\) are Fréchet differentiable on \(H\). This problem is unstable with respect to perturbations in the input data \(J(u)\), \(g_{i}\) and it is necessary to apply regularization to solve it. The authors propose and analyze a regularization method based on a gradient projection method of second order and a penalty function approach.