A PRIMAL-DUAL INTERIOR-POINT METHOD CAPABLE OF RAPIDLY DETECTING INFEASIBILITY FOR NONLINEAR PROGRAMS

被引:15
作者
Dai, Yu-Hong [1 ]
Liu, Xin-Wei [2 ]
Sun, Jie [2 ,3 ,4 ]
机构
[1] Chinese Acad Sci, Acad Math & Syst Sci, ICMSEC, LSEC, Beijing 100190, Peoples R China
[2] Hebei Univ Technol, Inst Math, Tianjin 300401, Peoples R China
[3] Curtin Univ, Sch Sci, Perth, WA, Australia
[4] Natl Univ Singapore, Sch Business, Singapore, Singapore
关键词
Nonlinear programming; constrained optimization; infeasibility; interior-point method; global and local convergence; GLOBAL CONVERGENCE; ALGORITHM; OPTIMIZATION;
D O I
10.3934/jimo.2018190
中图分类号
T [工业技术];
学科分类号
08 ;
摘要
With the help of a logarithmic barrier augmented Lagrangian function, we can obtain closed-form solutions of slack variables of logarithmic-barrier problems of nonlinear programs. As a result, a two-parameter primal-dual nonlinear system is proposed, which corresponds to the Karush-Kuhn-Tucker point and the infeasible stationary point of nonlinear programs, respectively, as one of two parameters vanishes. Based on this distinctive system, we present a primal-dual interior-point method capable of rapidly detecting infeasibility of nonlinear programs. The method generates interior-point iterates without truncation of the step. It is proved that our method converges to a Karush-Kuhn-Tucker point of the original problem as the barrier parameter tends to zero. Otherwise, the scaling parameter tends to zero, and the method converges to either an infeasible stationary point or a singular stationary point of the original problem. Moreover, our method has the capability to rapidly detect the infeasibility of the problem. Under suitable conditions, the method can be superlinearly or quadratically convergent to the Karush-Kuhn-Tucker point if the original problem is feasible, and it can be superlinearly or quadratically convergent to the infeasible stationary point when the problem is infeasible. Preliminary numerical results show that the method is efficient in solving some simple but hard problems, where the superlinear convergence to an infeasible stationary point is demonstrated when we solve two infeasible problems in the literature.
引用
收藏
页码:1009 / 1035
页数:27
相关论文
共 39 条
  • [1] Augmented Lagrangian methods under the constant positive linear dependence constraint qualification
    Andreani, R.
    Birgin, E. G.
    Martinez, J. M.
    Schuverdt, M. L.
    [J]. MATHEMATICAL PROGRAMMING, 2008, 111 (1-2) : 5 - 32
  • [2] A feasible BFGS interior point algorithm for solving convex minimization problems
    Armand, P
    Gilbert, JC
    Jan-Jégou, S
    [J]. SIAM JOURNAL ON OPTIMIZATION, 2000, 11 (01) : 199 - 222
  • [3] A local convergence property of primal-dual methods for nonlinear programming
    Armand, Paul
    Benoist, Joel
    [J]. MATHEMATICAL PROGRAMMING, 2008, 115 (02) : 199 - 222
  • [4] CUTE - CONSTRAINED AND UNCONSTRAINED TESTING ENVIRONMENT
    BONGARTZ, I
    CONN, AR
    GOULD, N
    TOINT, PL
    [J]. ACM TRANSACTIONS ON MATHEMATICAL SOFTWARE, 1995, 21 (01): : 123 - 160
  • [5] A SEQUENTIAL QUADRATIC OPTIMIZATION ALGORITHM WITH RAPID INFEASIBILITY DETECTION
    Burke, James V.
    Curtis, Frank E.
    Wang, Hao
    [J]. SIAM JOURNAL ON OPTIMIZATION, 2014, 24 (02) : 839 - 872
  • [6] A ROBUST SEQUENTIAL QUADRATIC-PROGRAMMING METHOD
    BURKE, JV
    HAN, SP
    [J]. MATHEMATICAL PROGRAMMING, 1989, 43 (03) : 277 - 303
  • [7] Byrd R. H., 1987, SIAM C OPT HOUST TX
  • [8] Byrd R. H., 1998, PITMAN RES NOTES MAT, V380, P37
  • [9] A trust region method based on interior point techniques for nonlinear programming
    Byrd, RH
    Gilbert, JC
    Nocedal, J
    [J]. MATHEMATICAL PROGRAMMING, 2000, 89 (01) : 149 - 185
  • [10] On the convergence of Newton iterations to non-stationary points
    Byrd, RH
    Marazzi, M
    Nocedal, J
    [J]. MATHEMATICAL PROGRAMMING, 2004, 99 (01) : 127 - 148