3 in the text Positive and Negative De64257nite Matrices and Optimization The following examples illustrate that in general it cannot easily be determined whether a sym metric matrix is positive de64257nite from inspection of the entries Example Cons ID: 74731
Download Pdf The PPT/PDF document "Jim Lambers MAT Summer Session Lecture..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Theprecedingexamplecanbegeneralizedasfollows:ifAisannndiagonalmatrixA=266664d1000d2.........000dn377775;thenAis:1.positivedeniteifandonlyifdi0fori=1;2;:::;n,2.negativedeniteifandonlyifdi0fori=1;2;:::;n,3.positivesemideniteifandonlyifdi0fori=1;2;:::;n,4.negativesemideniteifandonlyifdi0fori=1;2;:::;n,5.indeniteifandonlyifdi]TJ/;༕ ;.9; ;Tf 1;.51; 0 ;Td [;0forsomeindicesi,1in,andnegativeforotherindices.Wenowconsiderageneral22symmetricmatrixA=abbc:ThismatrixinducesthequadraticformQA(x;y)=ax2+2bxy+cy2:Ify=0,thenwehaveQA(x;0)=ax2,sowemustcertainlyhavea]TJ/;༕ ;.9; ;Tf 1;.51; 0 ;Td [;0inorderforAtobepositivedenite.Ify6=0,thenweuseachangeofvariablex=tyandobtainQA(x;y)=QA(ty;y)=at2y2+2bty2+cy2=(at2+2bt+c)y2:ThusQA(x;y)ispositivedeniteifandonlyif'(t)=at2+2bt+cispositive.From'0(t)=2at+2b;'00(t)=2a;weobtainthesinglecriticalpointt=b=aanddeterminethatitisastrictglobalminimizerof'(t).Theminimumvalueis'(t)=a(b=a)2+2b(b=a)+c=cb2 a=1 adet(A):WeconcludethatAispositivedeniteifandonlyifa0anddet(A)0.Thisleadstothefollowingtheorem.TheoremA22symmetricmatrixAis2 whichcanbeconrmedtobepositivedenitebyexamininationoftheprincipalminors:1=2,2=3,3=4.Itfollowsthatthecriticalpoint(0;0;0)isastrictglobalminimizeroff(x;y;z).2ExampleLetf(x;y)=exy+eyx:Wehaverf(x;y)=(exyeyx;exy+eyx);whichyieldsthecriticalpoints(x;x)forallx2R.WealsohaveHf(x;y)=exy+eyxexyeyxexyeyxexy+eyx;whichyields1=exy+eyx0;2=(exy+eyx)2(exyeyx)2=0:Thatis,Hf(x;x)ispositivesemidenite,making(x;x)aglobalminimizeroff(x;y).2ExampleLetf(x;y;z)=exy+eyx+ex2+z2:Wehaverf(x;y;z)=(exyeyx+2xex2;exy+eyx;2z)whichyieldstheconditionsz=0,x=yandx=0foracriticalpoint.Therefore,(0;0;0)istheonlycriticalpoint.WethenhaveHf(x;y;z)=24exy+eyx+4x2ex2+2ex2exyeyx0exyeyxexy+eyx000235;andthereforeHf(0;0;0)=2432022000235:whichispositivednite,inviewof1=3,2=2,and3=4.ItcanalsobeshownbydirectcomputationoftheminorsthatHf(x;y;z)ispositivedeniteonallofR3,usingthefactthatex0forallx.Therefore,thecriticalpoint(0;0;0)isastrictglobalminimizeroff(x;y;z).2ExampleLetf(x;y)=exy+ex+y:4 TheoremIff(x)isafunctionwithcontinuoussecondpartialderivativesonasetDRn,ifxisaninteriorpointofDthatisalsoacriticalpointoff(x),andifHf(x)isindenite,thenxisasaddlepointofx.ExampleLetf(x;y)=x312xy+8y3:Wehaverf(x;y)=(3x212y;12x+24y2);whichyieldsthecriticalpoints(0;0)and(2;1).WealsohaveHf(x;y)=6x121248x;andthereforeHf(0;0)=012120;Hf(2;1)=12121248:WeseethatHf(2;1)ispositivedenite,becauseitsprincipalminorsarepositive,butHf(0;0)isnot,as1=0and2=144.Thatis,Hf(0;0)isindenite,so(0;0)isasaddlepoint.Furthermore,limx!1f(x;y)=+1;limx!1f(x;y)=1;sof(x;y)hasnoglobalminimizeronR2.Wecanconclude,however,that(2;1)isastrictlocalminimizer.2ItshouldbeemphasizedthatiftheHessianispositivesemideniteornegativesemideniteatacriticalpoint,thenitcannotbeconcludedthatthecriticalpointisnecessarilyaminimizer,maximizerorsaddlepointofthefunction.ExampleLetf(x;y)=x4y4.Wehaverf(x;y)=(4x3;4y3);whichyieldsthecriticalpoint(0;0).WethenhaveHf(x;y)=12x20012y2:ThereforeHf(0;0)isthezeromatrix,whichispositivesemidenite.However,f(x;y)increasesfrom(0;0)alongthex-direction,anddecreasesalongthey-direction,so(0;0)isneitheralocalminimizernormaximizer.2ExampleLetf(x;y)=x4+y4.Asinthepreviousexample,Hf(0;0)isthezeromatrix,butitcanbeseenfromagraphthat(0;0)isastrictglobalminimizer.26