\(C^{k,1}\) functions, characterization, Taylor's formula and optimization: a survey (Q541377)

From MaRDI portal
scientific article
Language Label Description Also known as
English
\(C^{k,1}\) functions, characterization, Taylor's formula and optimization: a survey
scientific article

    Statements

    \(C^{k,1}\) functions, characterization, Taylor's formula and optimization: a survey (English)
    0 references
    0 references
    0 references
    0 references
    7 June 2011
    0 references
    The paper is a survey on \(C^{k,1}\) functions. The authors start with their own characterization of such functions by the uniform boundedness of some \((k+1)\)th order discretized differences, by the boundedness of the associated lower and upper \((k+1)\)th Riemann-type derivative. The \(C^{1,1}\) case is related to the known results with respect to Clarke's subdifferential and directional derivative. Then they cite Taylor's formula of \((k+1)\)th order -- some generalizations of Lebourg's mean value theorem -- using generalizations of Clarke's Jacobian for the \(k\)th derivative, cite Luc's implicit function theorem and give some classification of quasi convexity using some second order lower and upper Riemann-differentials (\(k=1\)). The optimization starts with necessary and sufficient optimality condition of \((k+1)\)th order using again \((k+1)\)th order lower and upper Riemann-differentials. Using the absolute continuity of the \(k\)th derivative in the sense of measure theory they prove a difference representation of order \(k\) and Taylor's formula of order \(k+1\), both in integral form for the remainder. They use these integral representations to estimate Lipschitz constant of the rest with respect to the increment. In the sequel necessary optimality conditions and new sufficient conditions higher order for strict minima arederived. As an illustration they discuss the cases \(k=0,1\). For constrained problems optimality conditions using the \((k+1)\)th order Dini set-valued derivative for \(k=0,1\) are presented. A convergence theorem of Sun of superlinear convergent Newton method for semi-smooth function is presented. A lot of results are already given in the monograph of \textit{D. Klatte} and \textit{B. Kummer} [Nonsmooth equations in optimization. Regularity, calculus, methods and applications. Nonconvex Optimization and Its Applications 60. Dordrecht: Kluwer Academic Publishers (2002; Zbl 1173.49300)] for \(C^{1,1}\) functions, numerics included. There are some misprints in the cited results and definitions and some not defined notations which hinder the readability.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    semi-smooth
    0 references
    Lipschitzian \(k\)th derivative
    0 references
    Taylor's formula
    0 references
    difference operators
    0 references
    nonsmooth optimization
    0 references
    sufficient and necessary optimality condition of higher order
    0 references
    Dini-derivative
    0 references
    Rieman-derivative
    0 references
    0 references