• No results found

Duality Theorems for Convex Semidefinite Optimization Problems with Conic Constraints

N/A
N/A
Protected

Academic year: 2020

Share "Duality Theorems for Convex Semidefinite Optimization Problems with Conic Constraints"

Copied!
13
0
0

Loading.... (view fulltext now)

Full text

(1)

Volume 2010, Article ID 363012,13pages doi:10.1155/2010/363012

Research Article

-Duality Theorems for Convex Semidefinite

Optimization Problems with Conic Constraints

Gue Myung Lee and Jae Hyoung Lee

Department of Applied Mathematics, Pukyong National University, Pusan 608-737, South Korea

Correspondence should be addressed to Gue Myung Lee,[email protected]

Received 30 October 2009; Accepted 10 December 2009

Academic Editor: Yeol Je Cho

Copyrightq2010 G. M. Lee and J. H. Lee. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

A convex semidefinite optimization problem with a conic constraint is considered. We formulate a Wolfe-type dual problem for the problem for its-approximate solutions, and then we prove

-weak duality theorem and-strong duality theorem which hold between the problem and its Wolfe type dual problem. Moreover, we give an example illustrating the duality theorems.

1. Introduction

Convex semidefinite optimization problem is to optimize an objective convex function over a linear matrix inequality. When the objective function is linear and the corresponding matrices are diagonal, this problem becomes a linear optimization problem.

For convex semidefinite optimization problem, Lagrangean duality without constraint qualification1,2, complete dual characterization conditions of solutions 1,3,4, saddle point theorems 5, and characterizations of optimal solution sets 6, 7 have been investigated.

To get the-approximate solution, many authors have established-optimality con-ditions,-saddle point theorems and-duality theorems for several kinds of optimization problems1,8–16.

Recently, Jeyakumar and Glover 11 gave -optimality conditions for convex optimization problems, which hold without any constraint qualification. Yokoyama and Shiraishi16gave a special case of convex optimization problem which satisfies-optimality conditions. Kim and Lee 12 proved sequential -saddle point theorems and -duality theorems for convex semidefinite optimization problems which have not conic constraints.

The purpose of this paper is to extend the -duality theorems by Kim and Lee

(2)

-weak duality theorem and -strong duality theorem for the problem and its Wolfe type dual problem, which hold under a weakened constraint qualification. Moreover, we give an example illustrating the duality theorems.

2. Preliminaries

Consider the following convex semidefinite optimization problem:

SDP Minimize fx,

subject to F0 m

i1

xiFi0, x1, x2, . . . , xmC,

2.1

where f : Rm → R is a convex function, C is a closed convex cone of Rm, and for i 0,1, . . . , m, FiSn, whereSn is the space of n× nreal symmetric matrices. The space Sn is partially ordered by the L ¨owner order, that is, forM, NSn, M Nif and only if MNis positive semidefinite. The inner product inSn is defined byM, N TrMN, where Tr·is the trace operation.

LetS:{MSn|M0}. ThenSis self-dual, that is,

S :θS

n|θ, Z0, for any ZSS. 2.2

Let Fx : F0 mi1xiFi,Fx :

m

i1xiFi, x x1, . . . , xm ∈ Rm.Then F is a linear operator fromRmtoSnand its dual is defined by

FZ TrF

1Z, . . . ,TrFmZ, 2.3 for anyZSn.Clearly,A:{xC|FxS}is the feasible set of SDP.

Definition 2.1. Letg:Rn → R∪ {∞}be a convex function.

1The subdifferential ofg ata ∈ domg,where domg {x ∈ Rn | gx < ∞}, is given by

∂ga v∈Rn|gxga v, xa , xRn, 2.4

where·,· is the scalar product onRn.

2The-subdifferential ofgata∈domgis given by

∂ga v∈Rn|gxga v, xa,x∈Rn. 2.5

Definition 2.2. Let 0.ThenxAis called an-approximate solution of SDP, if, for any xA,

(3)

Definition 2.3. The conjugate function of a functiong :Rn → R∪ {∞}is defined by

gv supv, xgx|xRn. 2.7

Definition 2.4. The epigraph of a functiong :Rn → R∪ {∞},epi g, is defined by

epigx, r∈Rn×R|gxr. 2.8

Ifgis sublineari.e., convex and positively homogeneous of degree one, then∂g0 ∂g0, for all 0. Ifgx gxk,x∈Rn,k ∈R, thenepigepig∗ 0, k. It is worth nothing that ifgis sublinear, then

epig∂g0×R

. 2.9

Moreover, ifgis sublinear and ifgx gxk,x∈Rn, andk∈R, then

epig∂g0×k,. 2.10

Definition 2.5. LetCbe a closed convex set inRnandxC.

1Let NCx {v ∈ Rn | v, yx 0,for allyC}. ThenNCx is called the normal cone toCatx.

2Let0. LetNCx {v∈Rn| v, yx ,for allyC}. ThenNCxis called the-normal set toCatx.

3When C is a closed convex cone in Rn, NC0 we denoted byC∗ and called the negative dual cone ofC.

Proposition 2.6see17,18. Letf :Rn → Rbe a convex function and letδC be the indicator function with respect to a closed convex subset C ofRn, that is,δCx 0ifxC, andδCxifx /C. Let0. Then

∂fδC x 00,10

01

∂0fx ∂1δCx.

2.11

Proposition 2.7see7. Letg : Rn → Rbe a continuous convex function and let h : Rn → R∪ {∞}be a proper lower semicontinuous convex function. Then

epighepigepih. 2.12

(4)

Lemma 2.8. LetFiSn, i 0,1, . . . , m. Suppose thatA /.Letu ∈ Rm and α ∈ R. Then the following are equivalent:

i

xC|F0 m

i1

Fixi 0

⊂ {x∈Rm| u, x α},

ii

u α

∈cl

Z,δS×R

FZ

−TrZF0−δ

C×R

.

2.13

3.

-Duality Theorem

Now we give -duality theorems for SDP. Using Lemma 2.8, we can obtain the following lemma which is useful in proving our-strong duality theorems for SDP.

Lemma 3.1. LetxA. Suppose that

Z,δS×R

FZ

−TrZF0δ

C×R

3.1

is closed. Thenxis an-approximate solution of SDP if and only if there existsZSsuch that for anyxC,

fx−TrZFxfx. 3.2

Proof. ⇒Letxbe an-approximate solution of SDP. Thenfxfx, for anyxA. Let hx fxfx . Then hx δAx 0, for anyx ∈ Rn. Thus we have, from

Proposition 2.7,

0∈epihδAepihepiδA

epif0, fx epiδA,

3.3

and hence,0, −fxepifepiδA. So there existsu, repif∗such that−u, fxrepiδA∗and hence there existsu, repif∗such that−u, x fxr for anyxA. Sincefur,−u, x fxfufor anyxA; and hence it follows fromLemma 2.8

that

u

fx fu

Z,δS×R

FZ

−TrZF0−δ

C×R

. 3.4

Thus there existZ, δS×R, c∗∈C∗, andγ∈Rsuch that

uFZc,

(5)

This gives

FZ, xc, xfx u, xfxfu

−TrZF0−δγfx ,

3.6

for anyx∈Rn. Thus we have

fxu, x fx−TrZF0−δγ

fxFZ, xc, xTrZF

0−δγ

fx−TrZFx c, xδγ

fx−TrZFx

3.7

for anyxC.

⇐Suppose that there existsZSsuch that

fx−TrZFxfx, 3.8

for anyxC. Then we have

fxfx−TrZFxfx, 3.9

for anyxA.Thusfxfx, for anyxA. Hencexis an-approximate solution of SDP.

Now we formulate the dual problem SDD of SDP as follows:

SDD maximize fx−TrZFx,

subject to 0∈∂0fxFZ NC1x, Z0,

01∈0, .

3.10

We prove-weak and-strong duality theorems which hold between SDP and SDD.

Theorem 3.2-weak duality. For any feasible solutionxof SDP and any feasible solutiony, Z of SDD,

(6)

Proof. Letxandy, Zbe feasible solutions of SDP and SDD respectively. Then TrZFx0 and there existv∂0fyandωN1Cysuch thatvωFZ. Thus, we have

fxfy −TrZFy v, xy0Tr

ZFy

ωFZ, xy 0Tr

ZFy

FZ, xy

0−1Tr

ZFy

FZ, xFZ, y 0−1

TrZFy

Tr

Zm i1

xiFi

−Tr

Zm i1

yiFi

0−1

TrZF0 Tr

Zm i1

yiFi

TrZFx0−1

0−1

.

3.12

Hencefxfy−TrZFy.

Theorem 3.3-strong duality. Suppose that

Z,δS×R

FZ

−TrZF0−δ

C×R

3.13

is closed. Ifx is an-approximate solution of SDP, then there existsZS such that x, Z is a 2-approximate solution of SDD.

Proof. LetxAbe an-approximate solution of SDP. Thenfxfx,for anyxA.By

Lemma 3.1, there existsZSsuch that

fx−TrZFxfx, 3.14

for anyxC. Lettingxxin3.14, TrZFx. SinceFxSandZS, TrZFx 0.

Thus from3.14,

(7)

for anyxC. Hencexis an-approximate solution of the following problem:

maximize fx−TrZFx,

subject to xC,

3.16

and so, 0 ∈ ∂fFZ δC x, and hence, byProposition 2.6, there exist0,1 ∈ 0, such that01and

0∈∂0fxF

ZN1Cx. 3.17

So,x, Zis a feasible solution of SDD. For any feasible solutiony, Zof SDD,

fx−TrZFxfy −TrZFy fxfy −TrZFy

−TrZFx

−TrZFx

by-weak duality

−2.

3.18

Thusx, Zis a 2-approximate solution to SDD.

Now we characterize the-normal set toRn.

Proposition 3.4. Letx1, . . . , xn∈Rnand0.Then

N

Rn

x1, . . . , xn

i0

n i1i

n

i1

Ai,

3.19

where

Ai

⎧ ⎪ ⎨ ⎪ ⎩

−R ifxi0,

$ −xi

i,0

%

ifxi>0.

(8)

Proof. Letx1, . . . , xn∈Rnand0. Then

N

Rn

x1, . . . , xn ∂δRnx1, . . . , xn

n

i1

δR×···×R×R×R×···×R

x1, . . . , xn

i0

n i1i

n

i1

∂iδR×···×R×R×R×···×Rx1, . . . , xn.

3.21

Letv1, . . . , vn∂iδR×···×R×R×R×···×Rx1, . . . , xn whereRis at theith position inR× · · · ×R×

R×R× · · · ×R

⇐⇒ for any y1, . . . , yn ∈R× · · · ×R×R×R× · · · ×R,

iv1

y1−x1 · · ·viyixi · · ·vnynxn ,

⇐⇒ for any yi∈R, iviyixi , vj0,

forj ∈ {1, . . . , n} \ {i},

⇐⇒ vi

⎧ ⎪ ⎪ ⎨ ⎪ ⎪ ⎩

−R, ifxi0,

$

i

xi,0

%

, ifxi>0,

vj0 for j∈ {1, . . . , n} \ {i}.

3.22

Thus, we have

N1Rn

x1, . . . , xn

i0

n i1i

n

i1

{0} × · · · × {0} ×Ai× {0} × · · · × {0}

i0

n i1i

n

i1 Ai.

3.23

FromProposition 3.4, we can calculateNR2

(9)

Corollary 3.5. Letx1, x2∈R2

and0.Then following hold.

iIfx1, x2 0,0, thenNR2

0,0 −R

2

.

iiIfx1, x2 x1,0andx1>0, thenNR2

x1,0 −/x1,0×−∞,0. iiiIfx1, x2 0, x2andx2>0, thenNR2

0, x2 −∞,0×−/x2,0. ivIfx1, x2 x1, x2,andx1>0andx2>0, then

N

R2

x1, x2

10,20, 12

$

1

x1 ,0

% ×

$

2

x2 ,0

%

. 3.24

Now we give an example illustrating our-duality theorems.

Example 3.6. Consider the following convex semidefinite program.

SDP Minimize x1x22,

subject to

0 x1 x1 0

0,

x1, x2∈R2.

3.25

Letfx1, x2 x1x22,

F0

0 0

0 0

, F1

0 1

1 0

, F2

0 0

0 0

, 3.26

and0. Letfx1, x2 x1x22and

Fx1, x2

0 x1 x1 0

. 3.27

ThenA:{0, x2∈R2|x20}is the set of all feasible solutions of SDP and the set of all -approximate solutions of SDP is{x1, x2∈R2|x1 0,0x2 }. LetF{x1, x2, Z| 0∈∂0fx1, x2−FZ NR12

x1, x2, Z 0, 01 ∈0, }. ThenFis the set of all feasible

(10)

A:

0,0,

a b b c

|0∈∂0f0,0−F

a b b c

NR12

0,0,

a0, c0, b2 ac,

01∈0,

0,0,

a b

b c

|0∈ {1} ×−2√0,2

0

−2b,0−R2,

a0, c0, b2 ac,

01∈0,

0,0,

a b b c

|2b,0∈−∞,1×−∞,2√0

,

a0, c0, b2 ac,

01∈0,

0,0,

a b

b c

|a0, c0, b1

2, b 2 ac

,

B:

0, x2,

a b b c

|x2>0, 0∈∂0f0, x2−F

a b b c N1 R2

0, x2

a0, c0, b2 ac, 010,

0, x2,

a b

b c

|x2>0, 0∈ {1} ×

2x2−2

0,2x22

0

−2b,0

−∞,

$

1

x2,0

%

, a0, c0, b2ac,

01∈0,

0, x2,

a b b c

|x2>0, 2b,0∈−∞,

$

2x2− 1 x2 −

2√0,2x22

0

%

,

a0, c0, b2ac,

01∈0,

0, x2,

a b

b c

|0< x2

0&021

2 , a0, c0, b 1 2, b

2ac,

01∈0,

(11)

C:

x1,0,

a b b c

|x1>0, 0∈∂0fx1,0−F

a b b c

NR12

x1,0,

a0, c0, b2 ac,

01∈0,

x1,0,

a b b c

|x1>0, 0∈ {1} ×

−2√0,2

0

−2b,0

$

1

x1,0

%

×−∞,0, a0, c0, b2ac,

01∈0,

x1,0,

a b b c

|x1>0, 2b,0∈

$

1− 1 x1

,1

%

×−∞,2√0

,

a0, c0, b2 ac,

01∈0,

x1,0,

a b b c

|0< x1,1−x12bx1, a0, c0, b 1 2, b

2ac,

01∈0,

,

D:

x1, x2,

a b b c

|x1>0, x2 >0, 0∈∂0fx1, x2−F

a b b c

NR12

x1, x2, a0, c0, b

2ac,

01∈0,

x1, x2,

a b b c

|x1>0, x2>0,

0∈ {1} ×2x2−2

0,2x22

0

−2b,0

11

x1 ,0

×

21

x2 ,0

,

a0, c0, b2 ac, 1

1121, 01∈0,

x1, x2,

a b b c

|x1>0, x2>0,

2b,0∈

1− 1 1 x1,1

×

2x2− 2

1 x2 −2

0,2x22

0

,

a0, c0, b2 ac, 1

1121, 01∈0,

(12)

x1, x2,

a b b c

|0< x1,11x12bx1,

0< x2

0

'

0221

2 , a0, c0, b 1 2, b

2ac, 1

1121, 01∈0,

.

3.28

ThusFABCD. We can check that for any x1, x2Aand anyy1, y2,a bb cF,

fx1, x2f

y1, y2 −Tr

a b b c

Fy1, y2

, 3.29

that is,-weak duality holds.

Letx1, x2 ∈Abe an-approximate solution of SDP. Thenx1 0 and 0x2

. So, we can easily check thatx1, x2,0 0

0 0 ∈F. Since Tr0 0

0 0 Fx1, x2 0, from3.29,

fx1, x2f

y1, y2 −Tr

a b b c

Fy1, y2

, 3.30

for anyy1, y2,

a b

b cF. Sox1, x2,

a b b c

is an-approximate solution of SDD. Hence -strong duality holds.

Acknowledgment

This work was supported by the Korea Science and Engineering FoundationKOSEFNRL Program grant funded by the Korean governmentMESTno. R0A-2008-000-20010-0.

References

1 V. Jeyakumar and N. Dinh, “Avoiding duality gaps in convex semidefinite programming without Slater’s condition,” Applied Mathematics Report AMR04/6, University of New South Wales, Sydney, Australia, 2004.

2 M. V. Ramana, L. Tunc¸el, and H. Wolkowicz, “Strong duality for semidefinite programming,”SIAM Journal on Optimization, vol. 7, no. 3, pp. 641–662, 1997.

3 V. Jeyakumar, G. M. Lee, and N. Dinh, “New sequential Lagrange multiplier conditions characterizing optimality without constraint qualification for convex programs,”SIAM Journal on Optimization, vol. 14, no. 2, pp. 534–547, 2003.

(13)

5 N. Dinh, V. Jeyakumar, and G. M. Lee, “Sequential Lagrangian conditions for convex programs with applications to semidefinite programming,”Journal of Optimization Theory and Applications, vol. 125, no. 1, pp. 85–112, 2005.

6 V. Jeyakumar, G. M. Lee, and N. Dinh, “Lagrange multiplier conditions characterizing the optimal solution sets of cone-constrained convex programs,”Journal of Optimization Theory and Applications, vol. 123, no. 1, pp. 83–103, 2004.

7 V. Jeyakumar, G. M. Lee, and N. Dinh, “Characterizations of solution sets of convex vector minimization problems,”European Journal of Operational Research, vol. 174, no. 3, pp. 1380–1395, 2006. 8 M. G. Govil and A. Mehra, “ε-optimality for multiobjective programming on a Banach space,”

European Journal of Operational Research, vol. 157, no. 1, pp. 106–112, 2004.

9 C. Guti´errez, B. Jim´enez, and V. Novo, “Multiplier rules and saddle-point theorems for Helbig’s approximate solutions in convex Pareto problems,”Journal of Global Optimization, vol. 32, no. 3, pp. 367–383, 2005.

10 A. Hamel, “Anε-lagrange multiplier rule for a mathematical programming problem on Banach spaces,”Optimization, vol. 49, no. 1-2, pp. 137–149, 2001.

11 V. Jeyakumar and B. M. Glover, “Characterizing global optimality for DC optimization problems under convex inequality constraints,”Journal of Global Optimization, vol. 8, no. 2, pp. 171–187, 1996. 12 G. S. Kim and G. M. Lee, “On ε-approximate solutions for convex semidefinite optimization

problems,”Taiwanese Journal of Mathematics, vol. 11, no. 3, pp. 765–784, 2007.

13 J. C. Liu, “ε-duality theorem of nondifferentiable nonconvex multiobjective programming,”Journal of Optimization Theory and Applications, vol. 69, no. 1, pp. 153–167, 1991.

14 J. C. Liu, “ε-Pareto optimality for nondifferentiable multiobjective programming via penalty function,”Journal of Mathematical Analysis and Applications, vol. 198, no. 1, pp. 248–261, 1996.

15 J.-J. Strodiot, V. H. Nguyen, and N. Heukemes, “ε-optimal solutions in nondifferentiable convex programming and some related questions,”Mathematical Programming, vol. 25, no. 3, pp. 307–328, 1983.

16 K. Yokoyama and S. Shiraishi, “Anε-optimal condition for convex programming problems without Slater’s constraint qualifications,” preprint.

17 J. B. Hiriart-Urruty and C. Lemarechal,Convex Analysis and Minimization Algorithms. I. Fundamentals, vol. 305 ofGrundlehren der mathematischen Wissenschaften, Springer, Berlin, Germany, 1993.

References

Related documents