Pseudopower expansion of solutions of generalized equations and constrained optimization problems (Q1906067)

This paper deals with high-order approximations of solutions of generalized equations of the form \(F(z, \varepsilon)+ N(z)\ni 0\), where \(F\in Z\times\mathbb{R}^+\to E\) and \(N\) is a multivalued operator with closed graph from \(Z\) to \(E\) and \(Z\) and \(E\) are Banach spaces. The authors use the technique of pseudopower expansion for the simple minimization problem \[ \min f(x, \varepsilon),\quad x\in K,\tag{P}{\(_\varepsilon\)} \] where \(X\) is a Banach space, \(f\) is a smooth function \(\in X\times\mathbb{R}^+\to\mathbb{R}\), and \(K\) is a closed convex subset of \(X\). Assuming that the problem (P)\(_0\) has a unique solution, the solution \(x^\varepsilon\) of the perturbed problem (P)\(_\varepsilon\) is computed in the form \(x^\varepsilon= x^0+ \varepsilon x^1(\varepsilon)+ \varepsilon^2 x^2(\varepsilon)+\cdots\;\). The optimality system of (P)\(_\varepsilon\) can be written as \(\nabla f(x^\varepsilon, \varepsilon)+ N_K(x^\varepsilon)\ni 0\), where \(\nabla\) denotes the gradient with respect to \(x\). \(N_K(x)\) is the set of outward normals to \(K\) at \(x\). An expansion of any order for the solution of a generalized equation of the form \[ F(z, \varepsilon)+ N(z)\ni 0 \] is justified under a strong regularity condition. The results are applied to the perturbation theory for constrained optimization problems.