NOT FOR QUOTATION WITHOUT PERMISSION OF THE AUTHOR
A NOTE ON THE &KERNEL FOE
QUASiDIPFEaSN'MABL6
FTJNCTIONSZ. Q. Xia
July, 1987 WP-87-66
Working Papers a r e interim r e p o r t s on work of t h e International Institute f o r Applied Systems Analysis and have received only lim- ited review. Views o r opinions expressed herein do not neces- sarily r e p r e s e n t those of t h e Institute or of i t s National Member Organizations.
INTERNATIONAL INSTITUTE FOR APPLIED SYSTEMS ANALYSIS A-2361 Laxenburg, Austria
The continuity of t h e star-kernel of quasidifferentials of a quasidif- ferentiable function with a star-equivalent bounded quasidifferential sub- family are studied and p a r t s of results are r e p r e s e n t e d in this note. I t also has been pointed out t h a t t h e directional subderivative and superderivative of a function can b e expressed as support functions of its star-kernel if t h e star-kernel c a n be generated by a quasidifferential of t h e function.
A. Kunhanski Program Leader System and Decision Sciences Program.
-
iii-
A NOTE ON THE @KERNEL FOR QUASIDIFFERENTLAB FUNCTIONS
2.
9.
X i aIn t h i s s h o r t n o t e t h e demonstrations of some propositions r e l a t e d t o t h e u p p e r semicontinuity of t h e S k e r n e l f o r quasidifferentiable functions a n d some examples concerning t h e S k e r n e l will be given, [I], [2] a n d [3].
Suppose t h a t f ( z ) is a quasidifferentiable function, defined on S
c
Rn where S i s a n open s e t , with a @-equivalent bounded quasidifferential subfamily. The notations we will use c a n be found o u t in [3]. Their definitions will not be r e p e a t e d h e r e .L.EIUMA 1
(a) u E a,f(z) - H ( z , u )
#
$ .(b) w E _ W ( u )
- v ( u
@ d ) Z < w , d > , V d € R n . (c) u E a , p ( z ) - u E ( Z , ~ ) .f ' ( z ; d ) S max <w , d >
-
w , u )=
6 ~I a,p(u
c ~ * ) ( o ) ) ,F o r t h e s a k e of convenience, assume t h a t &' i s a closed set.
LEMMA
2 If T ( Z , u C3 d ) is u p p e r semicontinuous in ( z , u ) E S Xa,
f ( z ) f o r e a c h d €IRn, t h e n t h e mapping a,f(*) andr(*
, *) are closed, i.e., u E a,f(z) and w EH(z
, u ) ( o r4 2
(u C3 *)(O)) whenever zi + x , ui -, u , w i -, w , a n d ui E a & ( z i ) , w i E ~ Y ( Z ~ , u i ) , i -,=.
PROOF Suppose zi -, z ,ui -, u , w i -, w and ui E
BJ
(zi ) , w i E _W(zi , ui ), i -,= .
According t o t h e L E M 1 (b), o n e h a sf o r e a c h d E Rn
.
S i n c e-
~ (, u C3 z d ) is u p p e r semicontinuous in ( z , u ) f o r e a c h d E R , o n e h a sIn o t h e r words, w E _W(Z , u )
.
I t follows from t h e L E M 1 (a) t h a tThe proof is completed. O
THEOREM 3. Suppose
DM
f ( z ) i s bounded uniformly in a neighborhood of z ,N, ( a ) ,
where 6 is a positive number. If ~ (, u C3 z d ) is u p p e r semicontinuous in ( z , u ) E S xa,
f ( z ) for e a c h d E Rn , t h e na,f
(*) is u p p e r semicontinuous.PROOF By contradiction, suppose
a d ( . )
i s not u p p e r semicontinuous. Then t h e r e e x i s t s a n o p e n set 0, a n E>
0 , a n d t h e r e e x i s t s sequences fzi1
a n d f u i1
such t h a tS i n c e
y DM
f ( z ) is bounded, t h e r e e x i s t s a s u b s e q u e n c etuik 1
c o n v e r g e n t2 E B ( O , E )
t o
u .
Theu
belongs t o OC b e c a u s e of 0 being a n o p e n set. Obviously,u
!Za,
f( z
). However, in t e r m s of t h e LEM. 2, we h a v es i n c e t h e mapping
a,
f ( * ) i s closed. This c o n t r a d i c t s t h e f a c t t h a tu
!Za,
f ( z ) . T h e r e f o r e , a , f ( * ) is a n u p p e r semi-continuous mapping. The t h e o r e m is p r o v e d . 3 LEXMA 4. S u p p o s e-
f' ( z
; d ) i s lower semi-continuous inz
E S f o r e a c h d E Rn a n d t h e mapping_W(z
,u )
is u p p e r semicontinuous in( z
,u )
E S X a , f ( z ) . Then t h e mapping a , f ( * ) i s closed.PROOF S u p p o s e t h a t
S i n c e f o r a n y
u
E a,f( z )
a n d f o r e a c h d E Rn ,' ( z )
max < w , d > ,-
w€ & ( I , U )
o n e h a s t h a t f o r
fzC 1
a n d{ui 1
t h e r e e x i s t s a s e q u e n c e lwC ] s u c h t h a tf '(zi ,
d C ) S max- <w
, d > ,w E l ( 2 1 , u )
According t o t h e assumption of t h e u p p e r semi-continuity of t h e mapping _W(. , *) t h e r e e x i s t s a s u b s e q u e n c e ,
{wik
{, c o n v e r g e n t t ow
s u c h t h a tw € _ W ( Z ,
u ) .
On t h e o t h e r hand, we h a v ef ' ( z , d ) S < w , d > ,
-
b e c a u s e of t h e lower semicontinuity of t h e function
-
f ' ( . , d ) . From t h i s , o n e h a s f ' ( z , d ) S max < w , d > .-
w € & ( I , U )Hence,
u
Ea , f ( z ) .
from LEY. 1 (d). The lemma h a s been p r o v e d .
THEOREM
5.
S u p p o s e f ' ( z , d ) i s lower semi-continuous inz
E S f o r e a c h d E R n , a n dDM
f ( z ) i s b k n d e d uniformly in a neighbourhood ofz
,N,
(6), a n d&(z
,u
) i s u p p e r semicontinuous in( z
,u
) E S x a e f ( z ) . Thenday(.)
i s a Gppersemicontinuous mapping.
THEOREM
6. If t h e r e e x i s t s a quasidifferentialLa,
Y ( z ) #5 , l ( z ) l
€ D M ,P(z) such t h a tand
then
a@s(z) =-a, s<z>
+zo s<z>
and
- -
a@s(z) = a, s ( z ) - a, s ( z ) ,
and
J ' ( z ; d )
=
max <u , d>
, V d E R~u @ @ f ( x ) and
PROOF. S i n c e V @ P ( z ) , z f ( z ) l € D M l ( z ) 6(*
1
B P ( z )+ Z P ( ~ ) )
2 6 ( .1 2,
J'(z> +So
J'(z))one h a s
~ ' ( 2 ; 0)s 6 ( .
1 3,
f ( z ) + Z o f ' ( z ) )-
In o t h e r words,
j"(z ; 0 )
=
6(.12,
f ( z )+ z,
f ( z ) ) .-
Theref o r e ,
a, s(z) + z,s(z)
ca,s(z).
According t o t h e definition of
3 ,
[3], w e have _u,(z> c_a, l ( z ) +z,
f ( z )S o t h e equality (3) holds. Similarly, (4) c a n b e p r o v e d in t h e same way. In t h i s case where t h e conditions (1) a n d (2) a r e s a t i s f i e d o n e h a s t h a t f o r a n y
u
Ea , f ( z )
a n d f o r a n y d E Rn t h e r e l a t i o n sq(u LL d )
=
-
u E e a f ( z ) max< u
,d >
and
-
q(u LLd)
=
max <u.
d>
2
' s @ f ( z )
are t r u e . Hence, t h e e q u a l i t i e s (5) a n d
( 6 )
are t r u e . The t h e o r e m h a s b e e n proved. I7EXAMPLE?
L e t f Ec'w").
Thena , f ( z ) = t v f ( z ) 1
,P f ( z ) = l o ] .
EXAMPLE 8 Let f b e a convex function defined in Rn
.
S u p p o s e 0,f ( z > = La, f ( z )
*[Oil
= [ a m )
,i o j i .
D f ( z > = L a f ( z )
#Z f ( z ) l
,w h e r e
a f ( z )
i s t h e s u b d i f f e r e n t i a l of f at z in t h e convex s e n s e . S i n c ea f ( z ) - - a f ( z ) = _ a p ( z ) ,
o n e h a s
a f ( z ) + Z P ( Z ) = a f ( z )
+( Z f ( z ) - - a m ) ) -
=@, 4 ( z ) + 5, f ( z ) )
+( Z f ( z ) - a f ( z ) )
a n d
- - -
a m ) - a f ( z ) = (5,
P ( Z )- a ,
~ ( 2 ) )+ ( Z J ( Z ) - 5 ~ ( ~ ) ) .
and
From Th. 6, o n e h a s
a , f ( z ) = a f ( z )
*P f ( z ) = to1
EXAMPLE
9. Let f b e a c o n c a v e function defined in Rn a n d Dof ( z ) = [I01 a, f ( z ) l
=[to{
D- a ( - f > ( z > l
and
I ) f ( z )
=
L a f ( z ) # Z f ( z ) l . Thena , f ( z )
=a, -
f ( z )=
- a ( - f ) ( z )and
- -
* f ( z )
= a, f ( ~ ) - a,
f ( z ) .EXAMPLE
10. Let f 1 be a convex function and f 2 b e a concave function defined in Rn , and f=
f 1+
f 2 . Then0, f ( z )
= Lao
f ( z ), 5 ,
f ( z ) l=
18 f l ( z ) 8 -8 ( - f 2 ) ( z)I
= D,
f l ( z ) ,- a,
f 2( z ) I .
For any
La
f ( z ) ,5
f ( z)I
E DM f ( z ) one h a s from (6) and Ex.9 t h a t BP(z) + Z . f ( z )=
C B f l ( z ) + Z f l ( z ) ) + caf,<z> + Z f 2 ( z ) )=
a f l ( z ) + ( B f l ( z )- -
O f ( z ) )+ w 2 ( z ) - 2 f 2 ( z ) )
-
a ( - f 2 ) ( z )=
( Z f l ( Z )-
Z f l ( 2 ) ) + CBP2(z) - B f 2 ( z ) ) + ( a f l ( z )-
a ( - f 2 ) ( z ) ) ,where
D f l ( z )
=
w l ( z ).
5 f l ( z ) l and9
D f 2 ( z >
=
W 2 ( z ) a f 2 ( z ) I.
I t follows from (8) t h a t
# ( z ) +
~ P ( z )
3 2 d l ( z ) + 5 d 2 ( z )=
apl(z)-
a ( - p 2 ) ( z ) . On t h e o t h e r hand, sinceand
9 9
8 f 2 ( z )
-
5 f 2 ( z )=
( ~ Q P ~ ( z )-
812(z))-
(&9'2(z) - 2 d 2 ( z ) ) , one h a s-
a f ( z )- Z J ( Z ) =
( Z f 1 ( z ) - 5 f l ( z ) ) + w 2 ( z ) - u 2 ( z ) ) + ( b d 2 ( z ) - 5 d 2 ( z ) ) .Hence
- - -
a f ( z )
- a f ( z )
33 d 2 ( z ) -
a d 2 ( z ) ) . From ( 9 ) , ( 1 0 ) and Th. 6 w e now havea d ( z ) = a f l ( z ) - a f 2 ( z )
= B d l ( z ) + Bd2(z)
and
-
a @ f ( z
)= 5,,.f2(z) - adZ(z 1) .
EXAMPLE
11Let f 1 and f 2 b e convex, and f=
f 1-
f 2 .T h e n w e h a v e
~ J ( z )
= a f i ( z ) - a f 2 ( z )
and
P ' ( z ) = a f 2 ( 2 ) - a f 2 ( z ) .
ACKNOWLEDGWENT
The a u t h o r i s deeply indebted t o Mrs. Adolfine Egleston f o r h e r g r e a t help.
REFERENCES
[ I ] Demyanov, V.F. a n d Rubinov, A.M. (1980). On Q u a s i d i f l e r e n t i a b L e f i n c t i o n - aLs Soviet Math. Dokl. Vol. 2, N o . 1 , pp. 14-17.
[ Z ] Demyanov, V.F. and Rubinov, A.M. (1985/1986). Q u a s i d i f l e r e n t i a L C a l c u l u s IIASA, Laxenburg Austria/ Springer-Verlag.
[ 3 ] Xia,