Solutions and optimality criteria for nonconvex quadratic-exponential minimization problem (Q929341)

From MaRDI portal
Revision as of 01:38, 5 March 2024 by Import240304020342 (talk | contribs) (Set profile property.)
scientific article
Language Label Description Also known as
English
Solutions and optimality criteria for nonconvex quadratic-exponential minimization problem
scientific article

    Statements

    Solutions and optimality criteria for nonconvex quadratic-exponential minimization problem (English)
    0 references
    0 references
    0 references
    17 June 2008
    0 references
    The authors consider the following nonconvex minimization problem: \[ ({\mathcal P}_{ e})\qquad \min\{P(x)=\tfrac12 x^T Ax-c^T x+W(x):\; x\in {\mathbb R}^n\}, \] where \(W\) is an exponential function with quadratic function exponent. The direct methods for solving the criticality condition \(\nabla P(x)=0\) are very difficult; moreover, the solutions can be only local, non global, minimizers. A potentially powerful methodology for solving global optimization problems is the so called {canonical duality theory}, and it is composed mainly of a canonical dual transformation and a triality theory. In this paper a concrete application of this theory is demonstrated by solving the nonconvex primal problem \({\mathcal P}_{ e}.\) The authors show that the nonlinear coupled algebraic system in \({\mathbb R}^n\) can be converted into an algebraic equation in a one-dimensional space via the canonical dual transformation, thereby obtaining a complete set of solutions. Then the triality theory allows to identify the set of local and global minimizers. In the end they illustrate a few examples where the theory presented in the paper is applied.
    0 references
    0 references
    duality theory
    0 references
    nonconvex programming
    0 references
    global optimization
    0 references
    quadratic-exponential function
    0 references
    nonlinear algebraic equation
    0 references
    triality
    0 references

    Identifiers