No label defined (Q3703595)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3930730
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 3930730

      Statements

      0 references
      1985
      0 references
      penalty functions
      0 references
      sequential quadratic programming
      0 references
      optimality criteria
      0 references
      superlinear convergence
      0 references
      Quasi-Newton methods
      0 references
      matrix update methods
      0 references
      local convergence
      0 references
      merit functions
      0 references
      active set strategies
      0 references
      [For the entire collection see Zbl 0566.00033.] The paper gives an overview (without proofs) on the theoretical background and on some practical aspects of sequential quadratic programming (SQP) methods for solving standard nonlinear programs \[ (P)\quad \min \{f(x) | g_ j(x)\leq 0\quad for\quad j=1,...,k;\quad g_ j(x)=0\quad for\quad j=k+1,...,m\}. \] Section 1 lists some known optimality criteria, and section 2 basics on the convergence of methods for solving nonlinear equations, among others the theorem of \textit{J. E. Dennis} and \textit{J. J. Moré} [SIAM Rev. 19, 46-89 (1977; Zbl 0356.65041)] characterizing the superlinear convergence of Quasi-Newton methods. Standard matrix update methods based on the least change secant update principle are considered in section 3, which also lists their convergence properties known so far. The central section 4 describes the local convergence properties of SQP-methods. It investigates their behaviour near Kuhn-Tucker points of (P), which are fixed points of the SQP-iteration, thus explaining why it is advisable to use update formulae which preserve positive definiteness. Moreover short new proofs of the Boggs-Tolle-Wang theorem (characterizing the one-step Q-superlinear convergence of SQP-methods) and of Powell's sufficient condition for the 2-step Q-superlinear convergence of these methods are given. Likewise the recent results of \textit{J. Nocedal} and \textit{M. Overton} [''Projected Hessian updating algorithms for nonlinearly constrained optimization'', Comp. Sci. Dept. Rep. No.95, Courant Inst., New York (1983)] on projected Hessian update algorithms are briefly explained. Several exact (differentiable and non-differentiable) penalty and merit functions, which serve to improve global convergence behaviour of SQP- methods, are described in section 5. The final section 6 deals with some aspects of the practical realization of SQP-methods (active set strategies, solution of quadratic subproblems, reduction to equality constrained quadratic programs etc.).
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references