lecture12

# lecture12 - Yinyu Ye MS&E Stanford MS&E211 Lecture...

This preview shows pages 1–5. Sign up to view the full content.

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document

This preview has intentionally blurred sections. Sign up to view the full version.

View Full Document
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #12 1 NonlinearOptimizationAlgorithmsIII YinyuYe DepartmentofManagementScienceandEngineering StanfordUniversity Stanford,CA94305,U.S.A. http://www.stanford.edu/˜yyye Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #12 2 LinearlyConstrainedConvexProblem ( LCCP ) minimize f ( x ) subjectto A x = b x ≥ . Weassumethat A hasfullrankand f isa differentiableconvex function. The KKTconditions : X s = A x = b- A T y + ∇ f ( x ) T- s = ( x , s ) ≥ . Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #12 3 BarrierFunction ( BLCCP ) minimize f ( x )- μ ∑ n j =1 log x j subjectto A x = b , whereparameter μ> isgiven.TheKKTconditionofthe barrieredproblem is ∇ f ( x ) T- μX- 1 e- A T y = A x = b x > , or X s = μ e A x = b- A T y + ∇ f ( x ) T- s = ( x , s ) > . Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #12 4 InteriorCentralPathsofLCCP Let ( x ( μ ) > , y ( μ ) , s ( μ ) > 0) bethe(unique)minimizerofthe barriered problem .Then,the centralpath ofLCCPcanbeexpressedas C = { ( x ( μ ) > , y ( μ ) , s ( μ ) > 0):0 <μ< ∞} . Theorem1 LetLCCPhave interiorfeasiblepoints andthebarrierfunctionis bounded forthegivendataset ( A, b ,f ) .Thenforany <μ< ∞ ,thecentral pathpoint ( x ( μ ) , y ( μ ) , s ( μ )) existsandisunique. Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #12 5 y a The objective hyperplanes Figure1:Thecentralpathina(dual)feasibleregion. Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #12 6 TheAnalyticCenter Inparticular,when μ = ∞ ,themaximizeofthe barrier functiononly: maximize ∑ n j =1 log x j subjectto A x = b , iscalledthe analyticcenter ofthefeasibleset. Theanalyticcenterforthefeasiblesetgivenasinequalities { y : A T y ≤ c } : maximize n X j =1 log s j , subjectto A T y + s = c , y free . maximize log( x 1 )+log( x 2 )+log( x 3 ) subjectto x 1 + x 2 + x 3 =1 Theanalyticcenteris x 1 = x 2 = x 3 =1 / 3 . Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #12 7 TheBarrierFunctionMethod Wehavea“good” approximate solutionfor ( BLCCP ) minimize f ( x )- μ ∑ n j =1 log...
View Full Document

{[ snackBarMessage ]}

### Page1 / 11

lecture12 - Yinyu Ye MS&E Stanford MS&E211 Lecture...

This preview shows document pages 1 - 5. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online