Mathematical foundations of nonsmooth embedding methods (Q805500)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Mathematical foundations of nonsmooth embedding methods
scientific article

    Statements

    Mathematical foundations of nonsmooth embedding methods (English)
    0 references
    0 references
    1990
    0 references
    This paper establishes some global homeomorphism results for nonsmooth functions of the form \(f=h\circ g\), where g is locally Lipschitzian and h has a property (strong Bouligand differentiability) generalizing the well-known property of strong Fréchet differentiability. A particular source of functions f having this composite form is the variational inequality problem in \({\mathbb{R}}^ n\), in which one is given a function F from an open subset \(\Omega\) of \({\mathbb{R}}^ n\) to \({\mathbb{R}}^ n\) and a closed convex subset C of \({\mathbb{R}}^ n\) meeting \(\Omega\). The problem then is to find some \(x_ 0\in C\cap \Omega\) such that for each \(x\in C\), \(<x-x_ 0,F(x_ 0)>\geq 0\) (if any such point exists). By writing \(\Pi_ C\) for the Euclidean projector on C, and \(f(x):=F(\Pi_ C(x))+(x-\Pi_ C(x))\), one obtains a (single-valued, nonsmooth) function f of the form just described, such that \(x_ 0\) solves the variational inequality problem if and only if \(f(x_ 0)=0\). In this way variational inequality problems can be converted into problems of finding zeros of single-valued functions, and therefore homeomorphism results for the latter are of obvious interest in establishing the existence of (and perhaps also for computing) solutions of the former. The author demonstrates an inverse-function theorem for functions of the form of f, and refers for more detail to his work in Math. Oper. Res. 16, No.2, 292-309 (1991). Then he gives two homeomorphism theorems, the first an application of the classical Palais theorem to the effect that a local homeomorphism is a global homeomorphism if and only if it is proper; here the inverse-function theorem is used to establish the local homeomorphism property. The second theorem extends the classical Hadamard theorem; it shows that if a certain generalized derivative of f (studied further in the paper cited above) has at every point \(x\in {\mathbb{R}}^ n\) an inverse that is locally Lipschitzian at the origin with modulus independent of x, then f is a homeomorphism of \({\mathbb{R}}^ n\) onto \({\mathbb{R}}^ n\).
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    classical embedding methods
    0 references
    continuation property
    0 references
    generalized equation
    0 references
    nonsmooth functions
    0 references
    strong Bouligand differentiability
    0 references
    variational inequality problem
    0 references
    inverse-function theorem
    0 references
    homeomorphism theorems
    0 references
    Hadamard theorem
    0 references