ON THE COMPLEXITY OF STEEPEST DESCENT, NEWTON'S AND REGULARIZED NEWTON'S METHODS FOR NONCONVEX UNCONSTRAINED OPTIMIZATION PROBLEMS

被引:159
作者
Cartis, C. [1 ]
Gould, N. I. M. [2 ]
Toint, Ph. L. [3 ]
机构
[1] Univ Edinburgh, Sch Math, Edinburgh EH9 3JZ, Midlothian, Scotland
[2] Rutherford Appleton Lab, Computat Sci & Engn Dept, Didcot OX11 0QX, Oxon, England
[3] Univ Namur, FUNDP, Dept Math, B-5000 Namur, Belgium
基金
英国工程与自然科学研究理事会;
关键词
nonlinear optimization; unconstrained optimization; steepest-descent method; Newton's method; trust-region methods; cubic regularization; global complexity bounds; global rate of convergence; CUBIC REGULARIZATION;
D O I
10.1137/090774100
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
It is shown that the steepest-descent and Newton's methods for unconstrained nonconvex optimization under standard assumptions may both require a number of iterations and function evaluations arbitrarily close to O(epsilon(-2)) to drive the norm of the gradient below epsilon. This shows that the upper bound of O(c(-2)) evaluations known for the steepest descent is tight and that Newton's method may be as slow as the steepest-descent method in the worst case. The improved evaluation complexity bound of O(epsilon(-3/2)) evaluations known for cubically regularized Newton's methods is also shown to be tight.
引用
收藏
页码:2833 / 2852
页数:20
相关论文
共 19 条
[1]  
[Anonymous], 1987, Unconstrained Optimization: Practical Methods of Optimization
[2]  
[Anonymous], 2004, INTRO LECT CONVEX OP
[3]  
Behforooz GH., 1995, Math Mag, V68, P289, DOI [10.2307/2690580, DOI 10.2307/2690580]
[4]  
CARTIS C, 2010, EFFICIENCY ADA UNPUB
[5]   Adaptive cubic regularisation methods for unconstrained optimization. Part II: worst-case function- and derivative-evaluation complexity [J].
Cartis, Coralia ;
Gould, Nicholas I. M. ;
Toint, Philippe L. .
MATHEMATICAL PROGRAMMING, 2011, 130 (02) :295-319
[6]   Adaptive cubic regularisation methods for unconstrained optimization. Part I: motivation, convergence and numerical results [J].
Cartis, Coralia ;
Gould, Nicholas I. M. ;
Toint, Philippe L. .
MATHEMATICAL PROGRAMMING, 2011, 127 (02) :245-295
[7]  
CONN AR, 2000, MPS SIAM SER OPTIM, V1
[8]  
DENNIS J. E., 1996, Numerical Methods for Unconstrained Optimization and Nonlinear Equations
[9]   A filter-trust-region method for unconstrained optimization [J].
Gould, NIM ;
Sainvitu, C ;
Toint, PL .
SIAM JOURNAL ON OPTIMIZATION, 2005, 16 (02) :341-357
[10]   Recursive trust-region methods for multiscale nonlinear optimization [J].
Gratton, Serge ;
Sartenaer, Annick ;
Toint, Philippe L. .
SIAM JOURNAL ON OPTIMIZATION, 2008, 19 (01) :414-444