See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/222128714

An improved definition of proper efficiency for vector

maximization with respect to cones

Article in Journal of Mathematical Analysis and Applications · September 1979

DOI: 10.1016/0022-247X(79)90226-9

CITATIONS READS

298 324

1 author:

Harold Benson
University of Florida
92 PUBLICATIONS 3,424 CITATIONS

SEE PROFILE

All content following this page was uploaded by Harold Benson on 30 June 2018.

The user has requested enhancement of the downloaded file.

JOURNAL OF MATHEMATICAL ANALYSIS AND APPLICATIONS 71, 232-241 (1979)

An Improved Definition of Proper Efficiency

for Vector Maximization with Respect to Cones

HAROLD P. BENSON

University of Florida, Gainesville, Florida 32611

Submitted by G. Leitntann

Recently Borwein has proposed a definition for extending Geoffrion’s concept

of proper efficiency to the vector maximization problem in which the domination
cone S is any nontrivial, closed convex cone. However, when S is the non-
negative orthant, solutions may exist which are proper according to Borwein’s
definition but improper by Geoffrion’s definition. As a result, when S is the
nonnegative orthant, certain properties of proper efficiency as defined by
Geoffrion do not hold under Borwein’s definition. To rectify this situation, we
propose a definition of proper efficiency for the case when S is a nontrivial,
closed convex cone which coincides with Geoffrion’s definition when S is the
nonnegative orthant. The proposed definition seems preferable to Borwein’s
for developing a theory of proper efficiency for the case when S is a nontrivial,
closed convex cone.

1. INTRODUCTION

Let S _C Rp be any nontrivial cone where p 3 2. Consider a vector-valued

criterion function

defined over a nonempty set XC R* where f,: X-t R Vj E J = {I, 2,...,p}.

For any t, u E R”, let t as u signify that t - u E S. Then the vector maximiza-
tion problem
VMAX:f(x) subject to x E X (P>

is the problem of finding all solutions that are efficient in the sense of the
following definition.

DEFINITION I .l . A point x is said to be an @cient solution of (P) if x E X

and f(x) >sf(~) for some x E X implies that f(x) = f(Z).
Adapting the terminology of Yu [Ill, S will be referred to as the domination
cone for (P). For the case when S = R+p, where R+P = {t E Rp / t, 2 0 Vj E J},
in order to eliminate anomalous solutions and to allow a more satisfactory
232
0022-247X/79/090232-10$02.00/0
Copyright 0 1979 by Academic Press, Inc.
All rights of reproduction in any form reserved.
 PROPER EFFICIENCY WITH RESPECT TO CONES 233

characterization, Geoffrion [S] has suggestedrestricting attention to efficient

solutions that are proper in the senseof the following definition. For any non-
empty setsA and B, let A/B denote {ZC1x E A, x $ B}.

DEFINITION 1.2 (Geoffrion). When S = R+B, a point 2 is a properly efficient

solution of (P) when it is an efficient solution of (P) and there exists a scalar
M > 0 such that for each i E J and each x EX satisfying fi(x) > f>(x), there
exists at leastonej E J/{;> withf,(x) <f,(g) and [f&x) - f@)]/[fi(~)-f,(x)] GM.
Geoffrion referred to an efficient point that is not properly efficient asimproperly
ejicient.
Recently Borwein [3] has proposed a definition for extending Definition 1.2
to the casewhen S is any nontrivial, closedconvex cone. When S = R+n),and
for all j E J, fj is a concave function on the convex set X, Borwein’s definition
coincideswith Geoffrion’s [3, p. 611.However, when S = R,” and at least one
objective function is nonconcave on X, solutions may exist which are proper
according to Borwein’s definition but improper by Geoffrion’s definition.
In this paper we propose a new definition for a properly efficient solution of
(P) when S is a nontrivial, closed convex cone. By strengthening Borwein’s
requirement for properness,this definition coincideswith Geoffrion’s definition
when S is the nonnegative orthant, even if some objective function is non-
concave. In Section 2 we present the new definition and explain how it
strengthens Borwein’s definition. In Section 3, we prove that when S = R+g,
our definition, unlike Borwein’s, coincideswith Geoffrion’s definition. We also
show the equivalence of our definition and Borwein’s under an appropriate
concavity assumption. In Section 4, properties of proper efficiency according
to our proposed definition and according to Borwein’s definition are compared.
Although these definitions yield identical extensionsof Geoffrion’s fundamental
results,when S = R,” all properties of proper efficiency asdefined by Geoffrion
hold under our proposeddefinition, but not under Borwein’s. From theseresults
we conclude, in Section 5, that our definition seemspreferable to Borwein’s
for developing a theory of proper efficiency when S is a nontrivial, closed
convex cone.

2. THE NEW DEFINITION OF PROPER EFFICIENCY

To extend Geoffrion’s definition of proper efficiency to the casewhen S is a

nontrivial, closed convex cone, Borwein used the tangent cone concept. Let
R, ={tERI t >Oo>.

DEFINITION 2.1. Let C C R” and w E C. The tangent coneto C at w, denoted

T(C, w), is the set of limits of the form h = lim h,(wi - w), where (h,) is a
sequencein R, and (wi} is a sequencein C with limit w.
234 HAROLD P. BENSON

Let f(X) = {f(x) j x EX} and, for any sets A, B 2 RJ, let il - B =
{a - b 1a EA and b EB). Borwein’s proposed definition is as follows.

DEFINITION 2.2 (Borwein). Let S C Rr be a nontrivial, closedconvex cone.

A point z is said to be a properly eficient solution of (P) when .c is an efficient
solution of (P) and T[f(X) - S,f(s)] n S = (01.
Our proposed definition of proper efficiency strengthens Borwein’s require-
ment for propernessso that when S = R +y, even if, for somei E J, f, is not
concave on X, the new definition and Geoffrion’s definition coincide. To
accomplishthis, the concept of a projecting cone will be used. For any set C,
let cl C denote the closure of C.

DEFINITION 2.3. Let CC RP. The projecting cone of C, denoted P(C), is

the set of all points h of the form h = ha, where h E R, and U’E C.
The projecting cone of a set C is alsoknown asthe conegenerated by C and
the conical extension of C. Canon, Cullum, and Polak [4], Kelley [6], Wijsman
[9, lo] and others have usedthe projecting cone concept.
Our proposed definition of a properly efficient solution is as follows.

DEFINITION 2.4. Let S C RP be a nontrivial, closedconvex cone. A point 3

is saidto be aproperly e@knt solution of(P) when I is an efficient solution of(P)
and cl P[f(X) - S - {f(s)}] n S = (0).
Following Geoffrion [Sj, an efficient solution which is not properly efficient
will be referred to as improperly eficient.
Let C _CRP and w E C. From Definitions 2.1 and 2.3, T(C, w) C cl P(C - {w}),
but the reversecontainment may not hold. Therefore, Definition 2.4 strengthens
Borwein’s requirement for proper efficiency by using the concept of a projecting
cone in place of the tangent cone concept. Geometrically, this stronger require-
ment can be interpreted asfollows. By considering only properly efficient solu-
tions as defined by Definition 2.4, any efficient solution f is excluded for which
sequences{yl} inf(X) - S and {h,} in R, exist such that the directions hi =
&[yi -f(x)] have a nonzero limit which belongsto S. Under Borwein’s require-
ment for proper efficiency, such a point x would not be excluded if the sequence
(y”} could not be chosenso asto converge tof(K).

3. EQUIVALENCE THEOREMS

Becauseof the strong requirement for propernessimposedby Definition 2.4,

this definition, unlike Borwein’s, coincides with Geoffrion’s definition of a
properly efficient solution of(P) when S = R, p, even if someobjective function
is nonconcave on X. Before proving this equivalence, consider the following
result.
 PROPER EFFICIENCY WITH RESPECT TO CONES 235
THEOREM 3.1. Let S C Rp be a nontrivial, closed convex cone. Suppose that f
is a concave function with respect to S on the convex set X.l Then Definitions 2.2
and 2.4 are equivalent.
Proof. Let K E X. Since S is a nontrivial, closed convex cone and f is concave
with respect to S on the convex set X, 5’ = f (X) - S is a convex set. From
Varaiya [7, Remark 2.11, cl P[V - (f (%)>I = T[V, f (a)] in this case.
From Theorem 3.1, the strengthening of Borwein’s requirement for proper-
ness embodied in Definition 2.4 does not alter his definition when f is concave
with respect to S on X. Therefore, in this case, all properties concerning proper
and improper efficiency that have been developed using Borwein’s definition
[l, 31 also hold under Definition 2.4.
We now prove the equivalence of Definition 2.4 and Geoffrion’s definition
when S = R&J).

THEOREM 3.2. When S = R+p, Definitions 1.2 and 2.4 are equivalent.
Proof. To prove the theorem, we will show that (i) if x is properly efficient
by Definition 1.2, then % is properly efficient by Definition 2.4 and that (ii)
the converse to (i) holds.
(i) The proof of this proceeds, with minor modifications, as does the
proof of a result shown earlier (see Borwein [3, Proposition 11) and will therefore
not be given.
(ii) Let % be an efficient solution of (P). Suppose that 2 is not properly
efficient according to Definition 1.2. Let {MJ be an unbounded sequence of
positive real numbers. Then, by reordering the objective functions, if necessary,
one can assume that for each A!& there exists an xi E X such that fi(xi) > fi(n) and

[fi(4 - fiwl/[f3w - f~W)l > Mz (3.1)

for all J’ E J/(1} such that f,(xi) < fj(%). By choosing a subsequence of {Mi},
if necessary, one can assume that

J = {i E J/U> Ifb”) < fX%)>

is constant for all i. (J” # @ by Definition 1.I.) For each i, let

xi = [fl(Xf) -f&v)]-1. (3.2)

Then Xi is positive for all i. Either (a) J/J = (1) or (b) j/l # {If.

1 A function f: X + RP is a concave function with respect to a cone S C RP on X, where

X !C Rn is a nonempty convex set, when, for any 01 such that 0 < 01 < 1 and for any
x1, x2 E x, f[cYxl + (1 - %)%“I - cxf (9) - (1 - CY)f (3”) E s.
236 HAROLD P. BENSON

Case 1. J/f = 11). For each j E J, consider

h, = lim h,[fj(xi) -f3(X)]. (3.3)

From (3.2), h, = 1.
Suppose j E f. From (3.1),

VW) - h@wlw) - fiw > 4Jw (3.4)

for all i. Since j E 1 and f,(xZ) > jr(a) for all i, we have that

0 > MW) - hbwch(~i) -ml (3.5)

for all i. Since {Mi} is an unbounded sequence of positive real numbers, (3.4)
and (3.5) together imply that

From (3.2) and (3.3), this implies that h, = 0. Therefore, hi = 0 Vj E g.

Since J/j = {l}, these results imply that

lim hJf(x”) -f(x)] E R+P/{O}.

According to Definition 2.4, 5 is not a properly efficient solution of (P). By the

contrapositive, the proof of (ii) for this case is complete.

Case 2. J/j # {l}. Suppose j E J/j, j # 1. Since j $ J” and f&+) > fi@)
for all i, we have that

for all i. Therefore, by choosing a suitable subsequence I, if necessary, the

limit given by

exists and is either a finite nonnegative number or + co. From (3.2), this implies
that

if one assumes that + co is an element of R, . Since any subsequence of a

convergent sequence converges to the same limit as the sequence, from the
proof for Case 1 it follows that
 PROPER EFFICIENCY WITH RESPECT TO CONES 237

for all j E g. From (3.2),

l& Wl(4 - ml = 1.
Therefore,

and z is not properly efficient for (P) according to Definition 2.4. By the contra-
positive, the proof of (ii) for this case is complete.
Prom Theorem 3.2, when S = R+P, and at least one objective function is
nonconcave on X, no solution which is improper according to Geoffrion’s
definition can be proper by Definition 2.4. However, such a solution may be
proper by Borwein’s definition. For instance, consider the following example.

EXAMPLE. Let S = R+2 and let f&r , x2) = xj Vj E J = {I, 2> be defined
on the nonconvex set X given by

X = [A,uA,uA,]~R+~,
where
4 = {(xl, ~2) E R2 I ~1 + x2 < 61,

A2 = {(xl, ~2) E R2 I ~1 d 21,

and
4 = {(xl, ~2) E R2 I x2 d 2).

By Definition 1.I, x = (3,3) is an efficient solution of the corresponding vector

maximization problem. For any N > 0, x = (2, N) E X. Therefore, there is
no M > 0 such that

[fib4 - .f2CWfi(~) - f&91 G M

for each x EX satisfyingf,(x) > f2(%). By G eoff rion’s definition, x is improperly

efficient. By Definition 2.3,

cl Pu(X) - S - {f(x)}] = B, u B, ,
where
Bj = {(A, , h,) E R2 I hi < 0}

Vj E J. Therefore, as required by Theorem 3.2, P is also improperly efficient

according to Definition 2.4. However, from Definition 2.1,

W(X) - W@)l = {@I , h2) E R2 I A, + h, < 01,

so that x is properly efficient according to Borwein’s definition.

238 HAROLD P. BENSON

4. COMPARISON OF PROPERTIES

Let us compare properties of proper efficiency according to Borwein’s defini-

tion with properties according to our proposed definition.
Using Definition 2.2, Borwein [3, Theorems 1 and 21hasextended Geoffrion’s
fundamental characterizationsof proper efficiency when 5’ = R+p [5, Theorems
1 and 21 to the casewhen S is a nontrivial, closedconvex cone. These extensions
also hold under our proposed definition of proper efficiency. In particular, we
have the following two results. Let (int S*) denote the interior of the dualcone S*
for S, where
S" = {s* ER" 1(s*,s) > Ob's~S}.

THEOREM 4.1. Assume S is a nontrivial, closed convex cone. Suppose that I

is optimal for
max(s*,f(x)) subject to x E X Ps*>

for somes* E (int S*). Then x is a properly eficient solution for (P) according to
DeJinition 2.4.

THEOREM 4.2. AssumeS is a nontrivial, closed convex cone. Suppose that

(int S*) # ia and that f is a concave function with respect to S on the convex
set X. Then f is a properly e#kient solution for (P) according to Definition 2.4 if
and only if 5 is optimal for (P$,) for some s* E (int S*).

The proof of Theorem 4.1 proceeds, with minor modifications, as does the
proof of Theorem 1 of Borwein [3]. Theorem 4.2 is an immediate consequence
of Borwein’s Theorem 2 [3] and our Theorem 3.1.
Although both Definition 2.4 and Borwein’s definition of proper efficiency
yield identical extensions of Geoffrion’s fundamental results, Definition 2.4
has a distinct advantage over Borwein’s definition. From Theorem 3.2, when
S = R+*, all properties of proper efficiency as defined by Geoffrion also hold
under Definition 2.4. However, certain properties of proper efficiency asdefined
by Geoffrion may not hold under Borwein’s definition when S = R+P and at
least one objective function is nonconcave on X. For example, consider the
following theorem. (This theorem has been shown previously only for the case
when, for all j E J, fi is a concave function on the convex set X [2, 81.)

THEOREM 4.3. Let S = R+* and assume that the problem

SUPc .fJ(4
J-3
subject to
(P,)
 PROPER EFFICIENCY WITH RESPECT TO CONES 239

and
XEX,

is unbounded for some 2 E X. Then, according to Definition 1.2, no properly e#cient

sohtiom of (P) exist.
Proof. Let f be an efficient solution of (P) and let (n/l,} be an unbounded
sequenceof positive real numbers. Since (P*) is unbounded, for each L > 0,
there exists an x E X such that

and

Therefore, by reordering the objective functions, if necessary, for each Mi

there exists an xi E X such that

fiW - fiC3 > Mi (4.1)

and
fjW 3 f&q e E I. (4.2)

By choosing a suitable subsequenceof (M,), if necessary,one can assumethat

3 = {.I+E J/W Ifdx’) < fml

is constant for all i. Supposej E J. Then from (4.2) and the definition of J,

fi@;) B fkq < fiW

for all i. Therefore, for eachj E J,

0 < fim - fM) G m - fm (4.3)

for all i.
Let M be any positive scalar. Pick some Mt, E (n/r,) such that Ml, > MB,
where B = max,&f@) - fj(Z)], Th en, f rom (4.1), fr(x”‘) -jr(%) > MLfj(x) -
fi(%)] Vj E p. Using (4.3) this implies that
fi(xi’) - f&?) > M[f@) - f&‘)] Vj E J.

Rearranging, we have that

VW) - fkWlfjW - fW)l > M Yi EJ.

Since M is an arbitrary positive scalar, x is not properly efficient according to
Definition 1.2.

409/71/I-16
240 HAROLD P. BENSON

Let S = R+“. Theorem 4.3 gives a sufficient condition for concluding that
no properly efficient solutions of (P) exist, according to Geoffrion’s definition.
However, if at least one objective function is nonconcave on X, this condition
is not sufficient to conclude that no properly efficient solutions of (P) exist
according to Borwein’s definition of proper efficiency. For example, in the
example in Section 3, the set of efficient points of the corresponding vector
maximization problem is {(xr , ~a) E R2 j x1 + x,=6and2<sI<4). The
problem
sup x1 + ss
subject to
Xl 2 0,
3 2 0,
and
(Xl > x2) E x

where X is as given in the example, is unbounded. As required by Theorem 4.3,

each efficient point is improperly efficient according to Geoffrion’s definition.
However, each efficient point is properly efficient by Borwein’s definition.

5. CONCLUSION

Using either Borwein’s definition or our proposed definition, Geoffrion’s

fundamental results characterizing proper efficiency can be extended to the
case when S is a nontrivial, closed convex cone. However, since our definition
contains a stronger requirement for properness than Borwein’s, our definition
coincides with Geoffrion’s definition when S = R,“. Therefore, when S = R+P,
all properties of proper efficiency as defined by Geoffrion also hold under our
proposed definition. Such properties may not hold under Borwein’s definition
when at least one objective function is nonconcave on X. Therefore, it seems
preferable to use our proposed definition instead of Borwein’s in developing a
theory of proper efficiency for the case when S is a nontrivial, closed convex cone.

REFERENCES

1. H. P. BENSON, “Efficiency and Proper Efficiency in Vector Maximization with

Respect to Cones,” Research Publication GMR-2557, General Motors Research
Laboratories, Warren, Mich., 1977.
2. H. P. BENSON, Existence of efficient solutions for vector maximization problems,
J. Optimization Theory Appl. 26 (1978), 569-580.
3. J. BORWEIN, Proper efficient points for maximizations with respect to cones, SIAM
J. Control Optimization 15 (1977), 57-63.
4. M. D. CANON, C. D. CULLUM, JR., AND E. POLAK, “Theory of Optimal Control
and Mathematical Programming,” McGraw-Hill, New York, 1970.
 PROPER EFFICIENCY WITH RESPECT TO CONES 241

5. A. M. GEOFFRION, Proper efficiency and the theory of vector maximization, J. Math.

Anal. Appl. 22 (1968), 618-630.
6. J. L. KELLEY et al., “Linear Topological Spaces,” Van Nostrand, Princeton, N. J.,
1963.
7. P. P. VARAIYA, Nonlinear programming in Banach space, SIAM J. Appl. Math.
15 (1967), 284-293.
8. R. E. WENDELL AND D. N. LEE, Efficiency in multiple objective optimization problems,
Math. Programming 12 (1977), 406-414.
9. R. A. WIJSMAN, Convergence of sequences of convex sets, cones and functions, Bull.
Amer. Math. Sot. 70 (1964), 186-188.
10. R. A. WIJSMAN, Convergence of sequences of convex sets, cones and functions, II,
Trans. Amer. Math. Sot. 123 (1966), 32-45.
11. P. L. Yu, Cone convexity, cone extreme points, and nondominated solutions in
decision problems with multiobjectives, J. Optimization Theory Appl. 14 (1974),
319-377.

View publication stats