Contents lists available atScienceDirect
Applied Numerical Mathematics
www.elsevier.com/locate/apnum
Unified smoothing functions for absolute value equation associated with second-order cone
Chieu Thanh Nguyen
a, B. Saheya
b,
1, Yu-Lin Chang
a, Jein-Shan Chen
a,∗,
2aDepartmentofMathematics,NationalTaiwanNormalUniversity,Taipei11677,Taiwan
bCollegeofMathematicalScience,InnerMongoliaNormalUniversity,Hohhot010022,InnerMongolia,PRChina
a r t i c l e i n f o a b s t r a c t
Articlehistory:
Received19February2018 Receivedinrevisedform12July2018 Accepted30August2018
Availableonline3September2018
Keywords:
Second-ordercone Absolutevalueequations SmoothingNewtonalgorithm
Inthispaper,weexploreaunifiedwaytoconstructsmoothingfunctionsforsolvingtheab- solutevalueequationassociatedwithsecond-ordercone(SOCAVE).Numericalcomparisons arepresented,whichillustratewhat kindsofsmoothingfunctions workwell alongwith thesmoothingNewtonalgorithm.Inparticular,thenumericalexperimentsshowthatthe wellknownlossfunctionwidelyusedinengineeringcommunityistheworstoneamong theconstructedsmoothingfunctions,whichindicatesthattheotherproposedsmoothing functionscanbeemployedforsolvingengineeringproblems.
©2018IMACS.PublishedbyElsevierB.V.Allrightsreserved.
1. Introduction
Recently,thepaper[36] investigatesafamilyofsmoothingfunctionsalongwithasmoothing-typealgorithmtotacklethe absolutevalueequationassociatedwithsecond-ordercone(SOCAVE)andshowstheefficiencyofsuchapproach.Motivated bythisarticle,wecontinuetoasktwonaturalquestions.(i)Whetherthereareothersuitablesmoothingfunctionsthatcan be employedforsolvingtheSOCAVE?(ii)IsthereaunifiedwaytoconstructsmoothingfunctionsforsolvingtheSOCAVE?
Inthispaper,weprovideaffirmativeanswersforthesetwoqueries.Inordertosmoothlyconveythestoryofhowwefigure outtheanswers,webeginwithrecallingwheretheSOCAVEcomesfrom.
Thestandardabsolutevalueequation(AVE)isintheformof
Ax
+
B|
x| =
b,
(1)where A
∈ R
n×n, B∈ R
n×n, B=
0,andb∈ R
n.Here|
x|
meansthecomponentwiseabsolutevalue ofvectorx∈ R
n.When B= −
I,whereI istheidentitymatrix,theAVE(1) reducestothespecialform:Ax
− |
x| =
b.
It isknownthat theAVE(1) wasfirst introducedby Rohnin [40],butwas termedby Mangasarian[34].Duringthe past decade,therehasbeenmanyresearcherspayingattentiontothisequation,forexample,Caccetta,QuandZhou[1],Huand
*
Correspondingauthor.E-mailaddresses:thanhchieu90@gmail.com(C.T. Nguyen),saheya@imnu.edu.cn(B. Saheya),ylchang@math.ntnu.edu.tw(Y.-L. Chang), jschen@math.ntnu.edu.tw(J.-S. Chen).
1 Theauthor’sworkissupportedbyNaturalScienceFoundationofInnerMongolia(AwardNumber:2017MS0125).
2 Theauthor’sworkissupportedbyMinistryofScienceandTechnology,Taiwan.
https://doi.org/10.1016/j.apnum.2018.08.019
0168-9274/©2018IMACS.PublishedbyElsevierB.V.Allrightsreserved.
Huang[12],JiangandZhang[19],KetabchiandMoosaei[20],Mangasarian[26–33],MangasarianandMeyer[34],Prokopyev [37],andRohn[42].
WeelaboratemoreaboutthedevelopmentsoftheAVE.MangasarianandMeyer[34] showthattheAVE(1) isequivalent to the bilinear program, the generalized LCP (linear complementarity problem), and to the standard LCP provided 1 is not an eigenvalue of A.Withthese equivalent reformulations,they also show that the AVE(1) is NP-hard in its general formandprovide existence results.Prokopyev [37] furtherimproves theabove equivalence whichindicates that theAVE (1) can be equivalently recast asLCP withoutany assumptionon A and B,andalso provides a relationship withmixed integer programming. Ingeneral, if solvable, the AVE(1) can have eitherunique solution ormultiple (e.g.,exponentially many) solutions.Indeed,various sufficiency conditionson solvabilityandnon-solvabilityofthe AVE(1) withunique and multiplesolutions arediscussedin [34,37,41].Some variantsofthe AVE,like theabsolutevalueequation associatedwith second-orderconeandtheabsolutevalueprograms,areinvestigatedin[14] and[45],respectively.
Recently, another type of absolutevalue equation, a natural extension ofthe standard AVE (1), is considered [14,35, 36].Morespecificallythefollowingabsolutevalueequation associatedwithsecond-ordercones,abbreviatedasSOCAVE,is studied:
Ax
+
B|
x| =
b,
(2)where A
,
B∈ R
n×n andb∈ R
n arethesame asthosein(1);|
x|
denotesthe absolutevalueof x coming fromthesquare root of the Jordan product “◦
” of x and x. What is the difference betweenthe standard AVE (1) and the SOCAVE (2)?Their mathematical formats look the same. In fact, the main difference is that
|
x|
in the standard AVE (1) means the componentwise|
xi|
ofeach xi∈ R
,i.e.,|
x| = (|
x1|, |
x2|, · · · , |
xn|)
T∈ R
n;however,|
x|
intheSOCAVE(2) denotesthevector satisfying√
x2
:= √
x
◦
x associatedwithsecond-orderconeunderJordanproduct.Tounderstanditsmeaning,we needto introducethedefinitionofsecond-ordercone(SOC).Thesecond-orderconeinR
n(
n≥
1)
,alsocalledtheLorentz cone,is definedasK
n:=
(
x1,
x2) ∈ R × R
n−1|
x2≤
x1,
where
·
denotestheEuclideannorm.Ifn=
1,thenK
nisthesetofnonnegativerealsR
+.Ingeneral,a generalsecond- orderconeK
couldbetheCartesianproductofSOCs,i.e.,K := K
n1× · · · × K
nr.
Forsimplicity,wefocusonthesingleSOC
K
nbecausealltheanalysiscanbecarriedovertothesettingofCartesianproduct.TheSOCisaspecialcaseofsymmetricconesandcanbeanalyzedunderJordanproduct,see[9].Inparticular,foranytwo vectorsx
= (
x1,
x2) ∈ R × R
n−1and y= (
y1,
y2) ∈ R × R
n−1,theJordanproduct ofx and y associatedwithK
nisdefinedasx
◦
y:=
xTy y1x2+
x1y2.
TheJordanproduct,unlikescalarormatrixmultiplication,isnotassociative,whichisamainsourceofcomplicationinthe analysisofoptimizationproblemsinvolvedSOC,see[5,6,10] andreferencesthereinformoredetails. Theidentityelement underthis Jordanproduct is e
= (
1,
0, ...,
0)
T∈ R
n. Withthesedefinitions, x2 means theJordan product ofx with itself, i.e.,x2:=
x◦
x;and√
x withx
∈ K
n denotestheuniquevectorsuchthat√
x◦ √
x
=
x.Inotherwords,thevector|
x|
inthe SOCAVE(2) iscomputedby|
x| := √
x◦
x.
Asremarkedintheliterature,thesignificanceoftheAVE(1) arisesfromthefactthattheAVEiscapableofformulating manyoptimizationproblemssuchaslinearprograms,quadraticprograms,bimatrixgames,andsoon.Likewise,theSOCAVE (2) plays a similar role in various optimization problemsinvolving second-order cones. There hasbeen manynumerical methodsproposed forsolving thestandardAVE(1) andtheSOCAVE(2).Please referto [36] fora quickreview.Basically, wefollowthesmoothingNewtonalgorithmemployedin[36] todealwiththeSOCAVE(2).Thiskindofalgorithmhasbeen apowerfultoolforsolvingmanyotheroptimizationproblems,includingsymmetricconecomplementarityproblems[21,23, 24],the systemofinequalitiesundertheorder inducedby symmetriccone [17,25,46], andso on.Itis alsoemployed for thestandard AVE(1) in [18,43]. Thenewupshotofthispaperliesondiscoveringmoresuitable smoothingfunctionsand exploringaunifiedwaytoconstructsmoothingfunctions.Ofcourse,thenumericalperformanceamongdifferentsmoothing functionsarecompared.Thesearetotallynewtotheliteratureandarethemaincontributionofthispaper.
Toclosethissection, werecall somebasicconcepts andbackgroundmaterials regardingthesecond-ordercone,which willbe used inthe subsequentanalysis. Moredetails can be foundin [5,6,9,10,14]. First, werecall theexpression ofthe spectraldecomposition ofx withrespecttoSOC.Forx
= (
x1,
x2) ∈ R × R
n−1,thespectraldecompositionofx withrespectto SOCisgivenbyx
= λ
1(
x)
u(x1)+ λ
2(
x)
u(x2),
(3)where
λ
i(
x) =
x1+ (−
1)
ix2fori=
1,
2 and u(xi)=
⎧ ⎪
⎨
⎪ ⎩
1 2
1
, ( −
1)
i xxT22
Tif
x2=
0,
1 2
1, ( −
1)
iω
TT if x2
=
0,
(4)
with
ω ∈ R
n−1 beinganyvector satisfyingω =
1.Thetwo scalarsλ
1(
x)
andλ
2(
x)
are calledspectral valuesof x;while thetwovectorsu(x1)andu(x2)arecalledthespectralvectorsofx.Moreover,itisobviousthatthespectraldecompositionof x∈ R
nisuniqueifx2=
0.Itisknownthatthespectralvaluesandspectralvectorspossesthefollowingproperties:(i) u(x1)
◦
u(x2)=
0 andu(xi)◦
u(xi)=
u(xi)fori=
1,
2;(ii)
u(x1)2=
u(x2)2=
12 andx2=
12(λ
21(
x) + λ
22(
x))
.Nextistheconceptabouttheprojectionontosecond-ordercone.Letx+denotetheprojectionofx onto
K
n,andx− be theprojectionof−
x ontothedualcone(K
n)
∗ ofK
n,wherethedualcone(K
n)
∗ isdefinedby(K
n)
∗:= {
y∈ R
n|
x,
y≥
0,
∀
x∈ K
n}
.Infact,thedualconeofK
n isitself,i.e.,( K
n)
∗= K
n.DuetothespecialstructureofK
n,theexplicitformulaof projectionofx= (
x1,
x2) ∈ R × R
n−1ontoK
n isobtainedin[5,6,8–10] asbelow:x+
=
⎧ ⎨
⎩
x if x
∈ K
n,
0 if x∈ − K
n,
u otherwise,
where u
=
⎡
⎣
x1+x2
2 x1+x22
x2x2
⎤
⎦ .
Similarly,theexpressionofx−canbewrittenoutas
x−
=
⎧ ⎨
⎩
0 if x
∈ K
n,
−
x if x∈ − K
n,
w otherwise,
where w
=
⎡
⎣ −
x1−2x2x1−x2 2
x2x2
⎤
⎦ .
Itiseasytoverifythatx
=
x++
x−andx+
= (λ
1(
x))
+u(x1)+ (λ
2(
x))
+u(x2) x−= (−λ
1(
x))
+u(x1)+ (−λ
2(
x))
+u(x2),
where
( α )
+=
max{
0, α }
forα ∈ R
.As fortheexpression of|
x|
associatedwithSOC.There is analternative wayvia the so-calledSOC-functiontoobtaintheexpressionof|
x|
,whichcanbefoundin[2,3].Inanycase,itcomesoutthat|
x| =
(λ
1(
x))
++ (−λ
1(
x))
+ u(x1)+
(λ
2(
x))
++ (−λ
2(
x))
+ u(x2)= λ
1(
x)
u(x1)+ λ
2(
x)
u(x2).
2. UnifiedsmoothingfunctionsforSOCAVE
AsmentionedinSection1,weemploythesmoothingNewtonmethodforsolvingtheSOCAVE(2),whichneedsasmooth- ingfunctiontoworkwith.Indeed,a familyofsmoothingfunctionswasalreadyconsideredin[36].Inthissection,welook into what kindsofsmoothingfunctionscan be employed towork withthe smoothingNewton algorithmfor solvingthe SOCAVE(2).
Definition2.1.Afunction
φ : R
++× R → R
iscalledasmoothingfunctionof|
t|
ifitsatisfiesthefollowing:(i)
φ
iscontinuouslydifferentiableat( μ ,
t) ∈ R
++× R
; (ii) limμ↓0
φ ( μ ,
t) = |
t|
foranyt∈ R
.Givenasmoothingfunction
φ
,wefurtherdefineavector-valuedfunction: R
++× R
n→ R
nas( μ ,
x) = φ ( μ , λ
1(
x))
u(x1)+ φ ( μ , λ
2(
x))
u(x2) (5)where
μ ∈ R
++isaparameter,λ
1(
x)
,λ
2(
x)
arethespectralvaluesofx,andu(x1),u(x2)arethespectral vectorsofx.Conse- quently,isalsosmoothon
R
++× R
n.Moreover,itiseasytoverifythatlim
μ→0+
( μ ,
x) = |λ
1(
x) |
u(x1)+ |λ
2(
x) |
u(x2)= |
x|
whichmeans eachfunction
( μ ,
x)
servesasa smoothingfunctionof|
x|
associatedwithSOC.Withthisobservation, for theSOCAVE(2),wefurtherdefinethefunction H( μ ,
x) : R
++× R
n→ R × R
n byH
( μ ,
x) =
μ
Ax
+
B( μ ,
x) −
b, ∀ μ ∈ R
++and x∈ R
n.
(6)Proposition2.1.Supposethatx
= (
x1,
x2) ∈ R × R
n−1hasthespectraldecompositionasin(3)–(4).LetH: R
++× R
n→ R
nbe definedasin(6).Then,(a) H
( μ ,
x) =
0 ifandonlyifxsolvestheSOCAVE(2);(b) H iscontinuouslydifferentiableat
( μ ,
x) ∈ R
++× R
nwiththeJacobianmatrixgivenby H( μ ,
x) =
1 0B∂(∂μμ,x) A
+
B∂(∂μx,x)(7)
where
∂( μ ,
x)
∂ μ =
∂φ ( μ , λ
1(
x))
∂ μ
u(1)
x
+ ∂φ ( μ , λ
2(
x))
∂ μ
u(2) x
,
∂( μ ,
x)
∂
x=
⎧ ⎪
⎪ ⎨
⎪ ⎪
⎩
∂φ(μ,x1)
∂x1 I if x2
=
0,
⎡
⎣
b cxT2
x2 cxx2
2 aI
+ (
b−
a)
xx2xT222
⎤
⎦
if x2=
0,
with
a
= φ ( μ , λ
2(
x)) − φ( μ , λ
1(
x)) λ
2(
x) − λ
1(
x) ,
b=
12
∂φ ( μ , λ
2(
x))
∂
x1+ ∂φ ( μ , λ
1(
x))
∂
x1,
(8)c
=
1 2∂φ ( μ , λ
2(
x))
∂
x1− ∂φ ( μ , λ
1(
x))
∂
x1.
Proof. (a)First,weobservethat
H
( μ ,
x) =
0⇐⇒ μ =
0 and Ax+
B( μ ,
x) −
b=
0⇐⇒
Ax+
B|
x| −
b=
0 andμ =
0.
Thisindicatesthatx isasolutiontotheSOCAVE(2) ifandonlyif
( μ ,
x)
isasolutiontoH( μ ,
x) =
0.(b)Since
( μ ,
x)
iscontinuously differentiable onR
++× R
n,itisclearthat H( μ ,
x)
iscontinuously differentiableonR
++× R
n.Thus,itremainstocomputetheJacobianmatrixofH( μ ,
x)
.Notethat( μ ,
x) = φ( μ , λ
1(
x))
u(x1)+ φ( μ , λ
2(
x))
u(x2)=
⎧ ⎪
⎪ ⎪
⎨
⎪ ⎪
⎪ ⎩
1 2φ ( μ , λ
1(
x)) + φ( μ , λ
2(
x))
−φ( μ , λ
1(
x))
xx2T2
+ φ( μ , λ
2(
x))
xxT22
if x2
=
0,
12
φ ( μ , λ
1(
x)) + φ( μ , λ
2(
x))
−φ( μ , λ
1(
x)) ω
T+ φ( μ , λ
2(
x)) ω
Tif x2
=
0=
1 2⎧ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎨
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎩
⎡
⎢ ⎢
⎢ ⎢
⎣
φ ( μ , λ
1(
x)) + φ( μ , λ
2(
x) (−φ( μ , λ
1(
x)) + φ( μ , λ
2(
x)))
xx¯22
.. .
(−φ( μ , λ
1(
x)) + φ( μ , λ
2(
x)))
x¯xn2
⎤
⎥ ⎥
⎥ ⎥
⎦
if x2=
0,
⎡
⎢ ⎢
⎢ ⎣
φ ( μ , λ
1(
x)) + φ( μ , λ
2(
x))
0.. .
0⎤
⎥ ⎥
⎥ ⎦
if x2=
0,
wherex2
:= (¯
x2, · · · , ¯
xn) ∈ R
n−1,ω = ( ω
2, · · · , ω
n) ∈ R
n−1.Fromchainrule,itistrivialthat∂ ( μ ,
x)
∂ μ =
∂φ ( μ , λ
1(
x))
∂ μ
u(1)
x
+ ∂φ ( μ , λ
2(
x))
∂ μ
u(2) x Inordertocompute ∂(μ,x)
∂x ,forsimplicity,wedenote
( μ ,
x) :=
1 2⎡
⎢ ⎢
⎢ ⎣ τ
1( μ ,
x) τ
2( μ ,
x)
.. . τ
n( μ ,
x)
⎤
⎥ ⎥
⎥ ⎦ .
Toproceed,wediscusstwocases.
(i)Forx2
=
0,wecompute∂ τ
1( μ ,
x)
∂
x1= ∂φ ( μ , λ
1(
x))
∂
x1+ ∂φ ( μ , λ
2(
x))
∂
x1= ∂φ ( μ , λ
1(
x))
∂λ
1(
x)
∂λ
1(
x)
∂
x1+ ∂φ ( μ , λ
2(
x))
∂λ
2(
x)
∂λ
2(
x)
∂
x1= ∂φ ( μ , λ
1(
x))
∂λ
1(
x) + ∂φ ( μ , λ
2(
x))
∂λ
2(
x) :=
2band
∂ τ
1( μ ,
x)
∂
x¯
i= ∂φ ( μ , λ
1(
x))
∂
x¯
i+ ∂φ ( μ , λ
2(
x))
∂
x¯
i= ∂φ ( μ , λ
1(
x))
∂λ
1(
x)
∂λ
1(
x)
∂
x¯
i+ ∂φ ( μ , λ
2(
x))
∂λ
2(
x)
∂λ
2(
x)
∂
x¯
i= − ∂φ ( μ , λ
1(
x))
∂λ
1(
x)
¯
xi x2+ ∂φ ( μ , λ
2(
x))
∂λ
2(
x)
¯
xi x2=
∂φ ( μ , λ
2(
x))
∂λ
2(
x) − ∂φ ( μ , λ
1(
x))
∂λ
1(
x)
x¯
i x2=
∂φ ( μ , λ
2(
x))
∂
x1− ∂φ ( μ , λ
1(
x))
∂
x1 x¯
i x2:=
2c x¯
i x2,
i=
2,· · · ,
n.
Moreover,
∂ τ
i( μ ,
x)
∂
x1=
∂φ ( μ , λ
2(
x))
∂
x1− ∂φ ( μ , λ
1(
x))
∂
x1 x¯
i x2=
2c¯
xi x2,
i=
2, · · · ,
n.
Similarly,wehave
∂ τ
2( μ ,
x)
∂
x¯
2=
∂φ ( μ , λ
2(
x))
∂
x¯
2− ∂φ ( μ , λ
1(
x))
∂ ¯
x2 x¯
2 x2+ (φ( μ , λ
2(
x)) − φ( μ , λ
1(
x)))
∂
x¯2x2
∂
x¯
2=
2b¯
x2· ¯
x2 x22+ (φ( μ , λ
2(
x)) − φ( μ , λ
1(
x)))
1 x2−
x¯
2· ¯
x2 x23=
2a+
2(
b−
a) ¯
x2· ¯
x2 x22,
wherea meansa
:= φ ( μ , λ
2(
x)) − φ( μ , λ
1(
x))
λ
2(
x) − λ
1(
x)
.Ingeneral,mimickingthesamederivationyields∂ τ
i( μ ,
x)
∂
x¯
j=
2a+
2(
b−
a)
x¯xi·¯xi22 if i
=
j,
2(
b−
a)
x¯xi·¯xj22 if i
=
j.
Tosumup,weobtain
∂ ( μ ,
x)
∂
x=
⎡
⎣
b cxT2
x2 cxx2
2 aI
+ (
b−
a)
x2x2Tx22
⎤
⎦
whichisthedesiredresult.
(ii)Forx2
=
0,itiscleartosee∂ τ
1( μ ,
x)
∂
x1=
2∂φ ( μ ,
x1)
∂
x1 and∂ τ
1( μ ,
x)
∂
x¯
i=
0 for i=
2,· · · ,
n.
Since
τ
i( μ ,
x) =
0 fori=
2, · · · ,
n,itgives ∂τi∂(xμ1,x)=
0.Moreover,∂ τ
2( μ ,
x)
∂ ¯
x2=
lim¯ x2→0
τ
2( μ ,
x1,
x¯
2,
0, · · · ,
0) − τ
2( μ ,
x1,
0, · · · ,
0)
¯
x2=
lim¯ x2→0
φ ( μ ,
x1+ |¯
x2|) − φ( μ ,
x1− |¯
x2|)
¯
x2¯
x2|¯
x2|
=
lim¯ x2→0
φ ( μ ,
x1+ |¯
x2|) − φ( μ ,
x1− |¯
x2|)
|¯
x2|
=
lim¯ x2→0
∂φ ( μ ,
x1+ |¯
x2|)
∂( |¯
x2|) − ∂φ ( μ ,
x1− |¯
x2|)
∂( |¯
x2|) (
as L’Hopital’s rule)
=
lim¯ x2→0
∂φ ( μ ,
x1+ |¯
x2|)
∂(
x1+ |¯
x2|) + ∂φ ( μ ,
x1− |¯
x2|)
∂(
x1− |¯
x2|)
=
2∂φ ( μ ,
x1)
∂
x1.
Thus,weobtain
∂ τ
i( μ ,
x)
∂
x¯
j=
2∂φ(∂μx,x1)1 if i
=
j,
0 if i
=
j,
whichisequivalenttosaying
∂( μ ,
x)
∂
x= ∂φ ( μ ,
x1)
∂
x1 I.
Fromalltheabove,weconcludethat
∂( μ ,
x)
∂
x=
⎧ ⎪
⎪ ⎨
⎪ ⎪
⎩
∂φ(μ,x1)
∂x1 I if x2
=
0,
⎡
⎣
b cxT2
x2 cxx2
2 aI
+ (
b−
a)
x2xT
x222
⎤
⎦
if x2=
0.
Thus,theproofiscomplete.
2
Now,we areready toanswerthe question aboutwhatkindofsmoothingfunctionscan beadopted inthesmoothing typealgorithm.Twotechnicallemmasareneededtowardstheanswer.
Lemma2.1.SupposethatM
,
N∈ R
n×n.Letσ
min(
M)
denotetheminimumsingularvalueofM,andσ
max(
N)
denotethemaximum singularvalueofN.Then,thefollowinghold.(a)
σ
min(
M) > σ
max(
N)
ifandonlyifσ
min(
MTM) > σ
max(
NTN)
. (b) Ifσ
min(
MTM) > σ
max(
NTN)
,thenMTM−
NTN ispositivedefinite.Proof. Theproofisstraightforwardorcanbefoundinusualtextbookofmatrixanalysis,soweomitithere.
2
Lemma2.2.LetA
,
S∈ R
n×nandA besymmetric.SupposethattheeigenvaluesofA andS STarearrangedinnon-increasingorder.Then,foreachk
=
1,
2, · · · ,
n,thereexistsanonnegativerealnumberθ
ksuchthatλ
min(
S ST) ≤ θ
k≤ λ
max(
S ST)
andλ
k(
S A ST) = θ
kλ
k(
A).
Proof. Pleasesee[11,Corollary4.5.11] foraproof.
2
We point out that the crucial key, which guarantees a smoothing function can be employed in the smoothing type algorithm,isthenonsingularity oftheJacobian matrix H
( μ ,
x))
givenin(7).Asbelow,we provideunderwhatcondition theJacobianmatrix H( μ ,
x))
isnonsingular.Theorem2.1.ConsideraSOCAVE(2) with
σ
min(
A) > σ
max(
B)
.LetH bedefinedasin(6).Supposethatφ : R
++× R → R
isa smoothingfunctionof|
t|
.If−
1≤
dtdφ ( μ ,
t) ≤
1 issatisfied,thentheJacobianmatrixH( μ ,
x)
isnonsingularforanyμ >
0.Proof. From the expression of H
( μ ,
x)
givenas in(7), we know that H( μ ,
x)
is nonsingularif andonly if the matrix A+
B∂(∂μx,x) isnonsingular.Thus,itsufficestoshowthatthematrix A+
B∂(∂μx,x) isnonsingularundertheconditions.Supposenot,thatis,thereexistsavector0
=
v∈ R
n suchthatA
+
B∂ ( μ ,
x)
∂
x v=
0whichimpliesthat
vTATA v
=
vT∂( μ ,
x)
∂
x TBTB
∂ ( μ ,
x)
∂
x v.
(9)Forconvenience,wedenoteC
:=
∂(∂μx,x).Then,itfollowsthat vTATA v=
vTCTBTBC v.ApplyingLemma2.2,thereexistsa constantˆθ
suchthatλ
min(
CTC) ≤ ˆθ ≤ λ
max(
CTC)
andλ
max(
CTBTBC) = ˆθλ
max(
BTB).
Notethatifwecanprovethat
0
≤ λ
min(
CTC) ≤ λ
max(
CTC) ≤
1,
we willhave
λ
max(
CTBTBC) ≤ λ
max(
BTB)
.Then,bytheassumptionthattheminimumsingular valueof A strictlyexceeds the maximumsingular value of B (i.e.,σ
min(
A) > σ
max(
B)
) and applyingLemma2.1,we obtain vTATA v>
vTCTBTBC v.Thiscontradictstheidentity(9),whichshowstheJacobianmatrix H
( μ ,
x)
isnonsingularforμ >
0.Thus, inlightoftheabovediscussion,itsuffices toclaim0
≤ λ
min(
CTC) ≤ λ
max(
CTC) ≤
1.Tothisend,we discusstwo cases.Case1: Forx2
=
0,wecomputethatC=
∂φ (∂μx1,x1)I.Since−
1≤
∂φ (∂μx1,x1)≤
1,itisclearthat0≤ λ(
CTC) ≤
1 forμ >
0.Then, theclaimisdone.Case2: Forx2
=
0,usingthefact thatthe matrixMTM isalways positivesemidefiniteforanymatrix M∈ R
m×n,we see thattheinequalityλ
min(
CTC) ≥
0 alwaysholds.Inordertoproveλ
max(
CTC) ≤
1,weneedtofurtherarguethatthematrix I−
CTC ispositivesemidefinite.First,wewriteoutI
−
CTC=
⎡
⎣
1−
b2−
c2−
2bcxx2T2−
2bcxx22
(
1−
a2)
I+ (
a2−
b2−
c2)
xx2x2T22
⎤
⎦ .
If
−
1<
∂φ (μ∂,λx1i(x))<
1,thenweobtain b2+
c2=
12
∂φ ( μ , λ
1(
x))
∂
x1 2+
∂φ ( μ , λ
2(
x))
∂
x1 2<
1.
Thisindicatesthat1
−
b2−
c2>
0.Byconsidering[
1−
b2−
c2]
asan1×
1 matrix,thissays[
1−
b2−
c2]
ispositivedefinite.Hence,itsSchurcomplementcanbecomputedasbelow:
(
1−
a2)
I+ (
a2−
b2−
c2)
x2xT 2
x22−
4b2c2 1−
b2−
c2x2xT2
x22= (
1−
a2)
I
−
x2x2T x22+
1
−
b2−
c2−
4b2c2 1−
b2−
c2 x2x2T x22.
(10)Ontheotherhand,bytheMeanValueTheorem,wehave
φ ( μ , λ
2(
x)) − φ( μ , λ
1(
x)) = ∂φ ( μ , ξ )
∂ξ (λ
2(
x) − λ
1(
x)),
where