## Analysis of Inexact Trust-Region Interior-Point SQP Algorithms (1995)

Citations: | 11 - 7 self |

### BibTeX

@TECHREPORT{Heinkenschloss95analysisof,

author = {Matthias Heinkenschloss and Luís N. Vicente},

title = {Analysis of Inexact Trust-Region Interior-Point SQP Algorithms},

institution = {},

year = {1995}

}

### Years of Citing Articles

### OpenURL

### Abstract

In this paper we analyze inexact trust-region interior-point (TRIP) sequential quadratic programming (SQP) algorithms for the solution of optimization problems with nonlinear equality constraints and simple bound constraints on some of the variables. Such problems arise in many engineering applications, in particular in optimal control problems with bounds on the control. The nonlinear constraints often come from the discretization of partial differential equations. In such cases the calculation of derivative information and the solution of linearized equations is expensive. Often, the solution of linear systems and derivatives are computed inexactly yielding nonzero residuals. This paper analyzes the effect of the inexactness onto the convergence of TRIP SQP and gives practical rules to control the size of the residuals of these inexact calculations. It is shown that if the size of the residuals is of the order of both the size of the constraints and the trust-region radius, t...

### Citations

1483 | GMRES a generalized minimal residual algorithm for solving nonsymmetric linear systems
- Saad, Schultz
(Show Context)
Citation Context ...und in [10], [47]. More recent references are [3], [4], [16], [17], [18], [27]. Most of the recent papers investigate the use of Krylov subspace methods for the solution of linear systems, like GMRES =-=[44]-=-, in inexact Newton methods. These Krylov subspace methods are attractive because they monitor the residual norm of the linear system in an efficient way and only require Jacobian times a vector, not ... |

257 |
Inexact Newton methods
- Dembo, Eisenstat, et al.
- 1982
(Show Context)
Citation Context ...bproblems. In the context of systems of nonlinear equations, inexact or truncated Newton methods have been proposed and analyzed by many authors. Some of the pioneer work in this area can be found in =-=[10]-=-, [47]. More recent references are [3], [4], [16], [17], [18], [27]. Most of the recent papers investigate the use of Krylov subspace methods for the solution of linear systems, like GMRES [44], in in... |

245 | Numerical Methods for Nonlinear Variational Problems - GLOWINSKI - 1984 |

222 |
An interior, trust region approach for nonlinear minimization subject to bounds
- Coleman, Li
- 1996
(Show Context)
Citation Context ...optimality conditions. In the following, we give a brief derivation of the form of the these conditions used in this paper. Further details can be found in [14]. For box constrained problems see also =-=[8]-=-, [15]. We introduce the Lagrangian function `(x; ) = f(x) +sT C(x) and we note that, due to the form of the bound constraints, the invertibility of C y (x) implies the linear independence of the acti... |

139 |
The conjugate gradient method and trust regions in large scale optimization
- Steihaug
- 1983
(Show Context)
Citation Context ...;s6 ffi k oe ; (3.9) wheres4 ,s5 , ands6 are positive constants independent of k. The use of conjugate gradients to solve trust--region subproblems in unconstrained minimization has been suggested in =-=[46]-=-, [48]. An adaptation of these algorithms to compute the tangential component that takes into account the problem structure and the bound constraints is presented in [14] and is given by: ALGORITHM 3.... |

112 | Representations of quasiNewton matrices and their use in limited memory methods
- BYRD, NOCEDAL, et al.
- 1994
(Show Context)
Citation Context ...(1.3). For both, the decoupled and the coupled approaches, we used approximations to reduced and to full Hessians. We approximate these matrices using the limited memory BFGS representations given in =-=[6]-=- with a memory size of 5 pairs of vectors. For the reduced Hessian we use a null--space secant update (see [41], [51]). The initial Hessian approximation is flI n\Gammam for the reduced Hessian and fl... |

99 |
Globally convergent inexact Newton methods
- EISENSTAT, WALKER
- 1994
(Show Context)
Citation Context ...ions, inexact or truncated Newton methods have been proposed and analyzed by many authors. Some of the pioneer work in this area can be found in [10], [47]. More recent references are [3], [4], [16], =-=[17]-=-, [18], [27]. Most of the recent papers investigate the use of Krylov subspace methods for the solution of linear systems, like GMRES [44], in inexact Newton methods. These Krylov subspace methods are... |

84 |
A numerical study of the limited memory BFGS method and the truncated-Newton method for large scale optimization
- Nash, Nocedal
- 1991
(Show Context)
Citation Context ...cit form. The results for the solution of systems of nonlinear equations have been extended to analyze inexact Newton methods for the solution of unconstrained minimization problems, e.g. [11], [37], =-=[39]-=-. In a recent paper [52], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by [2], [9], [12], [20... |

69 |
Newton-type minimization via the Lanczos method
- Nash
- 1984
(Show Context)
Citation Context ... explicit form. The results for the solution of systems of nonlinear equations have been extended to analyze inexact Newton methods for the solution of unconstrained minimization problems, e.g. [11], =-=[37]-=-, [39]. In a recent paper [52], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by [2], [9], [12... |

65 |
Truncated-Newton algorithms for large-scale unconstrained optimization
- Dembo, Steihaug
- 1983
(Show Context)
Citation Context ...ian in explicit form. The results for the solution of systems of nonlinear equations have been extended to analyze inexact Newton methods for the solution of unconstrained minimization problems, e.g. =-=[11]-=-, [37], [39]. In a recent paper [52], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by [2], [9... |

60 | Global convergence of a class of trust region algorithms for optimization with simple bounds
- CONN, GOULD, et al.
- 1988
(Show Context)
Citation Context ...1], [37], [39]. In a recent paper [52], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by [2], =-=[9]-=-, [12], [20], [32], [38], [50] among others. The papers [12], [20], [32], [38] investigating SQP methods mostly study the influence of inexactness on the local convergence rate. In [38] conditions on ... |

60 |
Quadratically constrained least squares and quadratic problems
- Golub, Matt
- 1991
(Show Context)
Citation Context ...mple, one can use the conjugate--gradient method applied to the normal equation as suggested for the general quadratic case in [46], [48], or one can use the Lanczos bidiagonalization as described in =-=[23]-=-. Both methods compute an approximate solution of (6.1) from a subspace that contains the negative gradient \GammaC y (x k ) T C k of the least squares functional. Thus, the steps s n k generated by t... |

54 |
Towards an efficient sparsity exploiting Newton method for minimization
- Toint
- 1981
(Show Context)
Citation Context ...i k oe ; (3.9) wheres4 ,s5 , ands6 are positive constants independent of k. The use of conjugate gradients to solve trust--region subproblems in unconstrained minimization has been suggested in [46], =-=[48]-=-. An adaptation of these algorithms to compute the tangential component that takes into account the problem structure and the bound constraints is presented in [14] and is given by: ALGORITHM 3.1 (EXA... |

52 | Convergence properties of a class of minimization algorithms, in Nonlinear Programming 2 - Powell - 1975 |

51 | Trust-Region Interior-Point Algorithms for Minimization Problems with Simple Bounds - Dennis, Vicente - 1996 |

47 | A global convergence theory for general trust-region-based algorithms for equality constrained optimization
- Dennis, El-Alem, et al.
- 1997
(Show Context)
Citation Context ...the inexactness that arises in the computation of directional derivatives of C(x). The details are presented in this paper. The convergence results presented in this paper rely on the theory given in =-=[13]-=-, [14]. A comprehensive convergence theory is presented in [49]. We have applied the inexact TRIP SQP algorithms to the solution of two optimal control problems, a boundary control problem for a nonli... |

42 | On the implementation of an algorithm for large-scale equality constrained optimization
- Lalee, Nocedal, et al.
- 1998
(Show Context)
Citation Context ...line search globalizations. Other papers which investigate SQP methods for large scale problems, but without treatment of inexact linear systems solves and inexact derivative information include [1], =-=[31]-=-, [36]. This paper is organized as follows. In Section 2, we state the first--order necessary optimality conditions of problem (1.1). A review of the features of the exact TRIP SQP algorithms necessar... |

38 | A reduced Hessian method for large-scale constrained optimization
- Biegler, Nocedal, et al.
- 1995
(Show Context)
Citation Context ... for line search globalizations. Other papers which investigate SQP methods for large scale problems, but without treatment of inexact linear systems solves and inexact derivative information include =-=[1]-=-, [31], [36]. This paper is organized as follows. In Section 2, we state the first--order necessary optimality conditions of problem (1.1). A review of the features of the exact TRIP SQP algorithms ne... |

37 | Trust-Region Interior-Point SQP Algorithms for a Class of Nonlinear Programming Problems
- Dennis, Heinkenschloss, et al.
- 1998
(Show Context)
Citation Context ... this and the continuity of D(x)W (x) T rf(x) we get lim i!1 ` k D k i W T k i rq k i (s n k i )k + kC k i k ' = 0: The rest of the proof is given in the last paragraph of the proof of Theorem 7.1 in =-=[14]-=-. The condition imposed in IA.4 is related to the computation of the reduced gradient. If the adjoint multipliers are used, then this condition can be interpreted as a restriction on how accurate thes... |

37 |
Projected Hessian updating algorithms for nonlinear constrained optimization calculation
- Nocedal, Overton
- 1985
(Show Context)
Citation Context ...ans. We approximate these matrices using the limited memory BFGS representations given in [6] with a memory size of 5 pairs of vectors. For the reduced Hessian we use a null--space secant update (see =-=[41]-=-, [51]). The initial Hessian approximation is flI n\Gammam for the reduced Hessian and flI n for the full Hessian, where fl is the regularization parameter in the objective function, set in both examp... |

34 | Interior point methods for optimal control of discrete time systems
- Wright
- 1993
(Show Context)
Citation Context ...aper [52], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by [2], [9], [12], [20], [32], [38], =-=[50]-=- among others. The papers [12], [20], [32], [38] investigating SQP methods mostly study the influence of inexactness on the local convergence rate. In [38] conditions on the inexactness are given that... |

33 |
Convergence theory of nonlinear Newton–Krylov algorithms
- Brown, Saad
- 1994
(Show Context)
Citation Context ...inear equations, inexact or truncated Newton methods have been proposed and analyzed by many authors. Some of the pioneer work in this area can be found in [10], [47]. More recent references are [3], =-=[4]-=-, [16], [17], [18], [27]. Most of the recent papers investigate the use of Krylov subspace methods for the solution of linear systems, like GMRES [44], in inexact Newton methods. These Krylov subspace... |

28 |
Continuation-conjugate gradient methods for the least squares solution of nonlinear boundary value problems
- Glowinski, Keller, et al.
- 1985
(Show Context)
Citation Context ... y , and y d = sin(2x 1 ) sin(2x 2 ). In this case the state equation (9.4) is a particular Bratu problem. Solvability and applications of the state equation are discussed e.g. in [21, Section IV.2], =-=[22]-=-. For the discretization of the problem, we use piecewise linear finite elements with a uniform triangulation obtained by first subdividing the x and the y subinterval into a sample of subintervals an... |

26 | e, Recent developments in algorithms and software for trust region methods - Mor' - 1983 |

25 |
A sequential quadratic programming algorithm using an incomplete solution of the subproblem
- Murray, Prieto
- 1995
(Show Context)
Citation Context ...earch globalizations. Other papers which investigate SQP methods for large scale problems, but without treatment of inexact linear systems solves and inexact derivative information include [1], [31], =-=[36]-=-. This paper is organized as follows. In Section 2, we state the first--order necessary optimality conditions of problem (1.1). A review of the features of the exact TRIP SQP algorithms necessary for ... |

24 |
A local convergence theory for combined inexact-Newton/finite-difference projection methods
- BROWN
- 1987
(Show Context)
Citation Context ... nonlinear equations, inexact or truncated Newton methods have been proposed and analyzed by many authors. Some of the pioneer work in this area can be found in [10], [47]. More recent references are =-=[3]-=-, [4], [16], [17], [18], [27]. Most of the recent papers investigate the use of Krylov subspace methods for the solution of linear systems, like GMRES [44], in inexact Newton methods. These Krylov sub... |

23 | A practical algorithm for general large scale nonlinear optimization problems
- Boggs, Kearsley, et al.
- 1999
(Show Context)
Citation Context ...g. [11], [37], [39]. In a recent paper [52], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by =-=[2]-=-, [9], [12], [20], [32], [38], [50] among others. The papers [12], [20], [32], [38] investigating SQP methods mostly study the influence of inexactness on the local convergence rate. In [38] condition... |

21 |
Minimization of a Large Scale Quadratic Function Subject to an Ellipsoidal Constraint. Department of Computational and Applied Mathematics
- Sorensen
- 1994
(Show Context)
Citation Context ...ntss2 ands3 , independent of k, such that kC k k 2 \Gamma kC y (x k )(s n k ) y + C k k 2s2 kC k k minf 3 kC k k; ffi k g: Another method to solve large scale trust--region subproblems is analyzed in =-=[45]-=-. In this approach the trust--region subproblem (6.1) is reformulated as an eigenvalue problem which is then solved by the implicitly restarted Lanczos method. This method and the method in [23] compu... |

20 | Global inexact Newton methods for very large scale nonlinear problems
- DEUFLHARD
- 1991
(Show Context)
Citation Context ... equations, inexact or truncated Newton methods have been proposed and analyzed by many authors. Some of the pioneer work in this area can be found in [10], [47]. More recent references are [3], [4], =-=[16]-=-, [17], [18], [27]. Most of the recent papers investigate the use of Krylov subspace methods for the solution of linear systems, like GMRES [44], in inexact Newton methods. These Krylov subspace metho... |

19 |
A global convergence theory for the Celis-Dennis-Tapia trust region algorithm for constrained optimization
- El-Alem
- 1991
(Show Context)
Citation Context ...rm q k (s k ; e k ) is given by q k (s k ; e k ) = ` k +rf T k s k +sT k (J k s k + e k ) + 1 2 s T k H k s k = q k (s k ) +sT k e k : (4.12) The update of the penalty parameter ae k follows El--Alem =-=[19]-=-. ALGORITHM 4.1 (INEXACT TRIP SQP ALGORITHMS). 1 Choose x 0 such that a ! u 0 ! b, pick ffi 0 ? 0, and calculates0 . Set ae \Gamma1s1 and ffl tol ? 0. Choose ff 1 , j 1 , oe, ffi min , ffi max , andsa... |

19 |
Projected sequential quadratic programming methods
- Heinkenschloss
- 1996
(Show Context)
Citation Context ...s that are simple to implement. To explain how we deal with inexactness and to present the main results of this paper, we need to introduce some of the structure of problem (1.1) (see also references =-=[24]-=-, [28], [29], [30]). For convenience we write x = / y u ! : Due to the partitioning of the variable x into y and u, the Jacobian matrix of C(x) can be written as J(x) = i C y (x) C u (x) j ; where C y... |

14 | An optimal control problem for flows with discontinuities
- Cliff, Heinkenschloss, et al.
- 1997
(Show Context)
Citation Context ...e numerical results are satisfactory and revealed interesting properties of the algorithms. Two examples are described in this section. Numerical results with two other examples are are documented in =-=[7]-=-, [25]. We use the formula (6.4) to compute the quasi--normal component, and conjugate-gradients to calculate the tangential component. The scheme used to update the trust radius and the inexact form ... |

9 |
An Infinite-Dimensional Convergence Theory for Reduced SQP Methods in Hilbert Space
- Kupfer
- 1996
(Show Context)
Citation Context ... are simple to implement. To explain how we deal with inexactness and to present the main results of this paper, we need to introduce some of the structure of problem (1.1) (see also references [24], =-=[28]-=-, [29], [30]). For convenience we write x = / y u ! : Due to the partitioning of the variable x into y and u, the Jacobian matrix of C(x) can be written as J(x) = i C y (x) C u (x) j ; where C y (x) 2... |

9 |
A prospective look at SQP methods for semilinear parabolic control problems, in Optimal control of partial differential equations
- Kupfer, Sachs
- 1991
(Show Context)
Citation Context ...imple to implement. To explain how we deal with inexactness and to present the main results of this paper, we need to introduce some of the structure of problem (1.1) (see also references [24], [28], =-=[29]-=-, [30]). For convenience we write x = / y u ! : Due to the partitioning of the variable x into y and u, the Jacobian matrix of C(x) can be written as J(x) = i C y (x) C u (x) j ; where C y (x) 2 IR m\... |

9 |
Numerical Solution of Parabolic State Constrained Control Problems usaing
- Leibfritz, Sachs
- 1994
(Show Context)
Citation Context ...n a recent paper [52], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by [2], [9], [12], [20], =-=[32]-=-, [38], [50] among others. The papers [12], [20], [32], [38] investigating SQP methods mostly study the influence of inexactness on the local convergence rate. In [38] conditions on the inexactness ar... |

8 |
Functional and Numerical Solution of a Control Problem Originating from Heat
- Burger, Pogu
- 1991
(Show Context)
Citation Context ...n we refer to [26]. 9.1. Boundary control of a nonlinear heat equation. The first example is the boundary control of a nonlinear heat equation. This and similar control problems are discussed e.g. in =-=[5]-=-, [30], [33], [40]. The goal is to control the heating process in such a way that the temperature at the boundarysx = 1 follows a certain desired temperature profile y d (t). The control u(t) acts on ... |

8 |
Optimal Control of Nonlinear Parabolic Systems
- aki, Tiba
- 1994
(Show Context)
Citation Context ...]. 9.1. Boundary control of a nonlinear heat equation. The first example is the boundary control of a nonlinear heat equation. This and similar control problems are discussed e.g. in [5], [30], [33], =-=[40]-=-. The goal is to control the heating process in such a way that the temperature at the boundarysx = 1 follows a certain desired temperature profile y d (t). The control u(t) acts on the boundary x = 0... |

6 | Recent developments in algorithms and software for trust region methods - E - 1983 |

6 |
Reduced Hessian algorithms for solving large-scale equality constrained optimization problems
- XIE
- 1991
(Show Context)
Citation Context ...e approximate these matrices using the limited memory BFGS representations given in [6] with a memory size of 5 pairs of vectors. For the reduced Hessian we use a null--space secant update (see [41], =-=[51]-=-). The initial Hessian approximation is flI n\Gammam for the reduced Hessian and flI n for the full Hessian, where fl is the regularization parameter in the objective function, set in both examples to... |

5 |
On inexact quasi–Newton methods for constrained optimization, in Numerical Optimization
- FONTECILLA
- 1984
(Show Context)
Citation Context ...39]. In a recent paper [52], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by [2], [9], [12], =-=[20]-=-, [32], [38], [50] among others. The papers [12], [20], [32], [38] investigating SQP methods mostly study the influence of inexactness on the local convergence rate. In [38] conditions on the inexactn... |

5 | A truncated SQP algorithm for large scale nonlinear programming problems
- Boggs, Tolle, et al.
- 1994
(Show Context)
Citation Context ...g. [12], [39], [41]. In a recent paper [55], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by =-=[2]-=-, [10], [13], [22], [34], [40], [53] among others. The papers [13], [22], [34], [40] investigating SQP methods mostly study the influence of inexactness on the local convergence rate. In [40] conditio... |

3 |
methods for the solution of optimal control problems governed by the Navier Stokes equations
- SQP
- 1996
(Show Context)
Citation Context ...erical results are satisfactory and revealed interesting properties of the algorithms. Two examples are described in this section. Numerical results with two other examples are are documented in [7], =-=[25]-=-. We use the formula (6.4) to compute the quasi--normal component, and conjugate-gradients to calculate the tangential component. The scheme used to update the trust radius and the inexact form of dia... |

3 |
solution of a nonlinear parabolic control problem by a reduced SQP method
- Numerical
- 1992
(Show Context)
Citation Context ...to implement. To explain how we deal with inexactness and to present the main results of this paper, we need to introduce some of the structure of problem (1.1) (see also references [24], [28], [29], =-=[30]-=-). For convenience we write x = / y u ! : Due to the partitioning of the variable x into y and u, the Jacobian matrix of C(x) can be written as J(x) = i C y (x) C u (x) j ; where C y (x) 2 IR m\Thetam... |

3 |
Conjugate residual methods for almost symmetric linear systems
- Meza, Symes
- 1992
(Show Context)
Citation Context ...nvestigated further. In particular the loss of symmetry due the inexactness deserves attention and the use of nonsymmetric methods for the solution of these subproblems will be investigated. See also =-=[34]-=-. In our applications the uniform boundedness of C y (x) \Gamma1 can be shown. However, the uniform boundedness of C y (x) \Gamma1 , or even the global invertibility is not guaranteed in other importa... |

2 |
Sequential truncated quadratic programming, in Numerical Optimization
- Dembo, Tulowitzki
- 1984
(Show Context)
Citation Context ...37], [39]. In a recent paper [52], the impact of inexactness in reduced gradient methods for design optimization has been analyzed. In nonlinear programming, inexactness has been studied by [2], [9], =-=[12]-=-, [20], [32], [38], [50] among others. The papers [12], [20], [32], [38] investigating SQP methods mostly study the influence of inexactness on the local convergence rate. In [38] conditions on the in... |

2 |
TRIP: A Package for the Solution of a Class of Constrained Optimization Problems; User's Guide
- Heinkenschloss, Vicente
- 1967
(Show Context)
Citation Context ...cation and the exposition of the importance of the scalar product. For a detailed description of how the scalar products are used and for a comprehensive description of the implementation we refer to =-=[26]-=-. 9.1. Boundary control of a nonlinear heat equation. The first example is the boundary control of a nonlinear heat equation. This and similar control problems are discussed e.g. in [5], [30], [33], [... |

2 |
Krylov--secant methods for solving systems of nonlinear equations
- IE, E, et al.
- 1995
(Show Context)
Citation Context ...t or truncated Newton methods have been proposed and analyzed by many authors. Some of the pioneer work in this area can be found in [10], [47]. More recent references are [3], [4], [16], [17], [18], =-=[27]-=-. Most of the recent papers investigate the use of Krylov subspace methods for the solution of linear systems, like GMRES [44], in inexact Newton methods. These Krylov subspace methods are attractive ... |

2 |
Optimal control of a CVD reactor for prescibed temperature behavior
- MEZA, PLANTENGA
- 1995
(Show Context)
Citation Context ...to [26]. 9.1. Boundary control of a nonlinear heat equation. The first example is the boundary control of a nonlinear heat equation. This and similar control problems are discussed e.g. in [5], [30], =-=[33]-=-, [40]. The goal is to control the heating process in such a way that the temperature at the boundarysx = 1 follows a certain desired temperature profile y d (t). The control u(t) acts on the boundary... |

2 |
and superlinear convergence for truncated iterated projections methods
- Local
- 1983
(Show Context)
Citation Context ...ems. In the context of systems of nonlinear equations, inexact or truncated Newton methods have been proposed and analyzed by many authors. Some of the pioneer work in this area can be found in [10], =-=[47]-=-. More recent references are [3], [4], [16], [17], [18], [27]. Most of the recent papers investigate the use of Krylov subspace methods for the solution of linear systems, like GMRES [44], in inexact ... |

2 |
Convergence theory of nonlinear Netwon–Krylov algorithms
- BROWN, SAAD
- 1994
(Show Context)
Citation Context ...inear equations, inexact or truncated Newton methods have been proposed and analyzed by many authors. Some of the pioneer work in this area can be found in [11], [50]. More recent references are [3], =-=[4]-=-, [18], [20], [19], [29]. Most of the recent papers investigate the use of Krylov subspace methods for the solution of linear systems, like GMRES [46], in inexact Newton methods. These Krylov subspace... |