• Nu S-Au Găsit Rezultate

View of Approximation of continuous set-valued functions in Fréchet spaces I

N/A
N/A
Protected

Academic year: 2022

Share "View of Approximation of continuous set-valued functions in Fréchet spaces I"

Copied!
5
0
0

Text complet

(1)

l4 I{üscyin Ror 4 }TA1'ËIE\{À'I-ÌCI\ --_ Tì I]VT-IE I)' ANÄI,Y S1ì N TJM í'R I QUE

I'T DI''I'TII1ORIIj ì)lj I,'T\I'T'IìOXIùIÀ'TION

T,"I.NÀI,YSIì NUNIÉRI{)UIì

ET

[,TI.

TIIEOITfiI ÐTì

I,IIPPROXTß'TATION Tome 20, No'

l-2'

1Í)$1,

pp. t5-23

tl- 1

ttt I

r-.1

x

:

O(1)

5

lA(ag.)

I.X, -l

O(r) m,þ,,,,X,,

: 0(l

) l) ÂP,

J

1) 1

+

O(1)

I,

I [3.*, I

N, -l û(I)

n,p,,,X,,,

: l)(1) as

?r¿

+

oo.

by virl,ne

of ,(1.õ), (2.1), (3.1),

and

(g.Z).

rt

_may be re,markerl

thai from

ìhe.'hJ.potheses

of the

theor,crn, (À,,)

is l¡o'ncletl. 1'his .a' be

sho\\'n

rike ilris."Si"";.

ii"j-is"non-de""náoir.,g_,

Nu

>-N_0,

which is a positive

constant.

rrence"i.rìq'i"rprio,. that

1i,1

ìí

bounded. Thus

APPROXIMATION Otr CONTINUOUS SET-VAI,UED F'UNCTIONS IN FRÉCHET SPACES. I-

i\IICIII]LIÌ I)\ù{I) i'I'I

(llari ) D:1

X

,il

t'1,

I

In,rl" : Iì

I À,1 l?,, lt-7t¿-tlt,l,,

:

O(t)

,Ð'

1r,,,Itt-1lt,,lr

ï\''e consiclor.scL-r'aluccl lunctions clcfinocl on a cornpacL Ilausdol'{f topological space and wìth conrpact colìvcti values in ¿r Fréchct spâce; \\¡c civc soltic conclilions rvhich ensttl'c th¿rt a sel

of sct-vaÌucd funcl"ions is a I{olovkil sc[ \'ith rospcc[ [o ccluicontinttous nets of nronotone Ì,iucar opelalors.

: o(1) i-ì' l^r,,,12, + o(t)

lÀ,,,

lx,,:t,(1) i.

þilxn

-l o(1)l),,,,1!*:o(l)

Ð,rs

rn

--+ oo,

in vien of (f.B),

(1.6), (2.1), ancl (8.2)

'Iherefore, 'rvc

get

i,hat

Xntrr¡duotion.

In

some r-ecenti pa,pers [2

],

[3,], 14] ancl

[6] ihe

interest

of l(orovkin-type approximation theory has been

extend.cd

to

linear

operators on

sct-r'alüed.

function

cones, mol,ivatecl

by the variei,y

of circumstances

in li'hich

set-valuecl

functions

are

involved,

such as

opti- rnal contlol theory, mathematical

ecoDomics

and probability theory.

In a precursory study of Yitale [6], the approximation of Ifaus-

cÌ.orff continuous sel-valued

functions rvil,h

compact convex values

in

a

finil,e

rlimensional normecl spa,ce has been cliscussecl

by

means

of

Bern-

"sl,ein polynomials

on

1,he oompact

rtal interval l0r1l.

Àfterwarcls,

Keimel anil

trìoth have establishec-l

in [3] a Korovkin-

h.1pe theorom

for

set-valuecl

llausclorff

conl,inuous

functions by

means of

Kotovkin positive

systems

for

single-valuecl

leal

functions

; the

same

theorem ha,s been imprr-rved

in [2] ]:¡. using suitable upper

a,nc1 lo$'et' trnvelopos.

Finally, I(eimel

and.

Iloth in [4]

have developecl

an

al¡st;r'act

formulation of

some

locally

convcx topologies on ordered cones

by intro-

ducing

a notion of

(upper, lorver)

continuity which constitutes the

sub-

stitute of the

trIausclorff

continuity in

normed spaces

; their

results gene-

n-alize approximation processes

to

1,he case

of

infinite-d.imensional

locally

convex

topological

spaces.

In this

paper we are interestetl

in

'bhe

Korovkin-type

approximation

of

set-valued" funct,ioirs

with

compact convex values

in n'réchei

spaces ;

Lhis

setting is not tho most

general

as

consid.ered.

in lal but

provides

u.s some tools

of

selecbion

theory which will

play a crucial rôle

in

achiev-

ing the main result.

Moreover,

the tr(orovkin-type

theorem obtained is rrob expressecl

in terms

of

l(olovkin

systems

for

single-valued.

real

funo-

fions

as

in

[a].

,i:',r+l'Jtn.,l":0

(1) as ??¿ -+ oo,

ror

t'

:

L¡2.

This

cornplel,os

the ploof of the

theorem

tìtìtltiÌìIìN0ljS

l' .ìl I c L t, 1'. i\I., Ott. rttt e .¡:l¿ttsion. oI nbsolulc stuttmabiliLg atrcl sr.lnc l]tcrtrents sf l,iltle¿:onl anrl PaIc¡¡, PIoc. Lo¡<ì. IIaLtr. Soc., Z (1957), 1.Ill-j.41.

2. o gl¡ c tl i irntT', Ìì., Sr.rl lcs sét'ies ul¡s¡tlumcnl sotiunablcs ptrr Ia mé.llrcrlc rles ntoqennr:s

aritlttnétirlttes, llull. Sci. l\,Iath., /a$ (1925), 234 - 256.

3'l\lislrra' Ii.N. ¿ìrÌdSrivasl-Íì\¡Ír,.|.ì.S.L.,Onrúsolule Oesùroiutnntcthility fru:laLs

of ittfirtitc scrrcs, Portuealiao ]\Iath., 4È (1983-.,1g84), 53- 61.

lìcccivcrì 15.IX.1990

l)eparlntcnl ol X,I0IlrcntaIit:s, Iìrcigcs U niaeNil!1, Iloyseri 38039,'l'Lu kcg

I Worli pcllortnctl under the auspiccs of thc G'N.A.F.r\' (C'N.R.) and M.IJ.R'S'T' t60o/.) and supportecl ìry I.N.d..A.nI.

(2)

Michelc Campiti

In the

second

part of this

paper, N'e

shall

consider singlevalued continuous functions

rvith

valuos

in a Fréchet

space artd

\re

shall stud5r

the

convergerìce

of

equicontinuous

nets of

opera,l,ors

satisf¡ri¡g

suitable

condil,iorx; as a

consequence, u'e

obtain a natulal

generalization oT the rvell-l<nou,n case

of

real

valued

functions.

l.

Prelirninulitt¡; ¿tutl

notirtiqrn. fn this

Secl,ion

lve recall sone

preli-

minary definitions

ancl some classical

lesults on

selc¡ction theor'¡'

l'hich v'ill be

usecl

in the

sequel.

AII the vectol

spa,ces

tnclel

consic-leral,ion

halre to

.lte consiclererl

over

1,he

lieid lR of real

numl¡els.

\\re

Ì:egin

liy

fi.xing a,

locally

con\¡cx

Ilausclorff

topological space

ancl

a

base

E of

cotLt'ex open neighborhoocls

of 0 in Z ;

rnoleovet', \\'e denote

by 6ont'(D) the cone of all non

ernpl,y convex subsel,s

of

-Ð, endon'ecl

rvith

1,he

natural

arlclil,ion ancl

rnultiplication by positive

scalars ancl

l¡v

(6Grnu(1tr)

the

se1,

of all tron ernpty

cornpact

convex

subsets ot E.

IrcL

'Y

ìte

a

topological sp¿ìce;

if

,f

:X

-Ør,,¿,,(-ãl)

is a

set-r.aluccl

function on

,Y, rve

recall thi¡i, a

(continuous) selection of

,/ is a

(con-

tinuous) function

<2:X

-,14

sal,isf¡'ing p(âr)

el'(r) for all r eX.

Nloreovor,

the Tollol'ing notations

r.r'ill be

ust¡ful in the

sequel ;

G(X, n)

denot;es l,he space

of all

corrl,inuous

functions on -T l'ith

values

in

-/J enrlou'ed

u,ith the

topologv

of the unifolm

con\¡ergence;

if ./r,...,1,, are set-r'alued Íunctions on N,

u,e consicler

the set-valuecl

functjon co(f

,...,f,) clefinetlb¡.puttins, fol

eß1rr ü

e-T,

co

(lr,..r,f ,)(t) :

co (lr(a),

"...,,f"(u)), I'here

ao(,ft(u),

...r1,@))

clenotes

the corn'exhulloli/t(ø)

u

...

...uÍ"(r);

rnoreorler,

if

cp

e(€(X,r), [ç]

clenotes the set-valuecl tunction defirred.bvputl,ing,

foleach

u

eX, {q}(r) : {ç(r)}

ancl

finall¡,,il|qy...¡9,, arein Ø(X)n)j

u'e sùnìrh'rvrit,e

co(9r,...,?,)

insteacl

of

co

({9r],...,[q,]).

the

follorving

plelirninary result is rvell-l<nol'n

ancl c¿rn

be

easily derit.ecl

from

lS, Ierlrnlzt

4.1 and lheoleur

3.2" ].

llr¿olost:r'rox 1.1.

If X is

(, palu,conlpad [fa,usd,ortf slta,ce

antl

IX

is

a

Frëchet sp&ce1 tlten, eu,cl¡ lozoer se'nt,icotttittttous set-au[,uedJu,n,ctiott ccchnits u co'ntinuous s election,,

Moreouer,

if

.f

:X

--+ GØr,ttu(14)'is ø lo¿oet' senticontitt,u,otts set-aa'lued Jwnction and

if A is a

closed subset

of X and 7:tl

--+

ll is

q, cott.tinuous ol ,f

ø,

th,en,

7

can be eutend,ed Lo ct, cotttinu,ous selectíon o,f

f ;

in,

', i."Í'noeX

eutcl

lloel@ò,

tltere

erists ø

continuous selection of

f

es l,he ao,lu,e yo

at no.

ffil

LeL

X

lre a palelcornpact Haustlor'llf space anû

Ð a Ilrí¡chct

space.

ff f :X --

G6c,¡¿t,(D)

is a

lorn'er semicontinuous set-v¿r,luecl

function, ìn

the sequel

s'e

shall denote

by 9t/(f) the

(non

emptv)

convex set of

all

con-

tinuous

selections

of /; bv viri,uo of the

preceding

Proposition 1.1

'rr'e have,

for

each a

eX,

(1.1)

T@)

-.¡¿",,{,v(n)}

.

Jlloreover,

the follorving

lernrna

ivill be useful in the following

Sections.

,4,pproxiìììation oI colrtint-tor¡s set-valtrecl frrncljons. I t7

þ'récl¡et sp&ce.

r.f

.f ,tt

:x -

8(€o)ruetr)

aie

loruer sem,icäinfilntol.ts set-,.ccr,l,ued

J'unct'iort

nnrl

), e [R,

il

resu,Its

(1.2) Yef(l -l tJ):9¿/(h

1-

e,,/(o),

(1.3) ?cr(xJ) : t,r"t$).

_

l]roo,[. Vtrehave,onl.ytoshorvtheinclusion.g"f (.f

-le)cgat(f)lgcf(cr), tht¡ otlrel

bein.g

trir.ial, 'ro this

enc-[,

fìx

9 e s,,cî(j'¡'¡1i'ancl let'(v',,¡,,.^'tr"q1

a

sequerce

of

convex closerl sets

in È l'hich

is a'úasé'of neighbòr,hó;ìi*

"i 0

"satisfS'ing trr,t,t

c 2-"V, fcir

e¿lch ra e

[\. I3y incluction,'1,e

shorv 1,he

existence^of a, se,quenoe

(ù,)"u^

oll coni,inuous selections of

,/ ând

a fìoq¡erlce (X¿),€sr ol' .onLi nuorrs scleclions

of

r¡ suclr Lhn,l

, frr,

r,acli .c e .t',

(1.4) {,n,(ø)

e tf ,,(r.')

*

l/

u,

x,. t(n)

e T,,eì)+ll

,, and

(l.ri)

rp(a)

* þ,(n)

y-,,(n)

ê

l; ,,.

th 'l {,'"V,lr'

pr

exist

Q.,, e

an

llorvs the

hen the funct,ions

tJ,,o :

-ll * il

ancl

üo(ø)

:=,fi,

1,,{")

,1,,{r), n@) :

,1,,,,r,(*)

x,{o)

fol

each n e

x,

are conl,jnuous selections of ,l' ancl r,espectivel¡, g. rlroleover

r{.,0

and

1e

satisfl

conilition (1.5).

ditions (l .,t) anrt (l

.i)

anrl con$idcr.

ing argunr.ent

to

shorv

the

existence aÛd Xn.t,t

of 9,+t satisf¡'ing ç(n) -

tT; ä{*l'n" derinitions

or

J"'

anct

_.By

(1.4-),

the

sequerrces

(ü,),,rm ancl

(2,),,e

or satisfy the

Cauchy eondil,ion

rvith

lespect

to

the r.rnifbrm

topology'il

zaçx,

n¡ änd

ilrerefore

us

funcl,ions qr: .Y

-- D

and, respec-

d X are

conl,inuous selections of

/

(1.5)).

shall

consicler. set-r.aluecl fnncl,ions

us an(l we shall indicate

these set-

(3)

18 Micl.rele Campiti

-valued functions simply as

continuous funcl,iolìs.

llhe

cone

of all

con-

tinnons

set-valued

funciions from a topological

space

X in €G"nu(fr) rvill

be clenoted

by G(X,

Øßouu(H)).

the

cone

G(X,

6(6r,t,u(D))

is

cquippecl

with the

l,opology

of

the

ulliform

convergence; namely we shall say

that a net (/,)Í.t of

set-valued lunctions

from,X

t'o G€onu(D) converges

to

a seb-valuecl

function /:X -'

-GGortr.,(H), if, for

eâch

I/ eE, there

exists øe

I

such

that

Observe

that if

l\foreoverr an

operator

'rt

:6 --

re¡,r,

gre'i."(t;jjîcallecr

*on',rr,nrìr, ;¡:

(2.1) 1,g e'€, ,f <

U =>

lt(fl.<

?,(t)

(cf.

( 1.7)),

, - Fina]j¡r,

\\'c s¿lv

that ?

is conl,inuous

if it

is oontinuous r.vith r,éìs'ìer,û

to

¡,he unit'olrrL syrrirrre.ûric toporogy irrduced

;,; Ø; ìïì;;, ii,:'åi,,r,"ìriii

thcrc

e.rists U e öt

*uclr tttat,'it"f,,"!.G;ti*ii, " '

u¡¡(¡

.l'(n)

-

s@)

I U, s(u) c f(n) |

U

for

each ,r

e,Y

(i.e.

,f < tl ï

(1,

g </+ U), ilren ,I(h

@)

-

T(s)

(r) I

tr,, ,I(s) (¿)

-

"(,fX

a)

a-

V.

for each-¿.e-{ (i.e. ff

1( is ¿ù rìor) errrpl,¡,

?(/) <

conrþacl ,.f\g)

* conv,ri 1,,

,f,(g)

subst,['ôf'lr, <

1'(./,) F

lr).

n,n rlerrol,c l¡r, 1,,..

Lhc

crnstanb

sef;-r,atuerr

furctior on .r' ,f .";ri*¿'i*rul i?';il :rìrjíï

laJ'

that

a subcone'z,of

Ø(x,l?Gont(E))

contains

iü"

"å".t*nt

sct_v¿rl*ecl

ftrnctions lhe ij.ft,

follorÀ'ins

e.e

Lernma

for

each

rvi[ '.o'

be

"rrìpty

irsörr¿

"orr,pu"t in^the ;;;.;;; ,ãq"ãi.

subset

K

or H.

Lnr'r'ri^ 2.1.

Let G

be u' suttcou,e

of G(x,

GG,,t,u(rx)) cortt*i,rt.i'tt¡1 il.e cct't¿stttrtt t¡tonotot¿e set-aa[,u,eù, Iin.eur oncÌ'rL!,.ort.J't.,ont .fu,nct,ions cut,il ,6 in,6(,1' Let

(?,)X.

be ct,n equ,ico,u,ti,ttrtotts t¿el, of

,

Çrß,.,,r,1b¡¡.'fnir,,,¡:oi:

,rrit,

i:'årl¿',

lltcre

erists

ü

e

S

sucl¿ lhî|,

l, I eG,.f 4

g

j-

U

.- I,ff) < ?',(t) |

V

.for

eacl¿ u

eI.

Proof

. l..'l, lI

e

tl ;

since

(?,)L, is

ecf

icontinurus at 0, ilre'e

exisl;s {.I,

eE

srrötr btrat, r'nr

å*"t-f ¿}r:'.'''

'

./ <

At

= I,(Í) ç Ì/ fol

each ¿ e 1.

+ i":åì'1?'.r

u-

co of Z(cf.,

e.g. l_

1' ant

set-valùed

..

In l4l,I(eirnel

and Roth have al ro inl,roclucetl

the

class

of uniformly

continuous operators; zr,ncl opelatot,

r'

: G

-G(x,GGnn"(n¡;r

called uni"-

l'ormly_gontinuous

(or briefly t¿-continuous)'rf, for u,ùrr" z et],

the;re

exists I/

e

E

such

that, for

eâch .f

,

g er€,

f <

s

* u- ?(/) < 'r(s)-lv.

4

al A¡rploxinratiol.r of contirtuous set-valnr:rl functious. I II

J'@)

-

"f,(c) -F

V,l,@) - l@) a ì;'fol

each n

e,\l

ancl ¿

)ø.

fn

[4,],

I(cinrel

and.

Iloth have

inl,roclucecl 1,he symrnetlic topology

on

(o,rru(fi))

by

considering

the farnily

({1)

eqo,',,@)

I

B - A + lr, A c B -f

tr/})r,Es

irs

a

neiglrbolhood base

of an arbitlary

element A. eGonu(E).

Sinc,e rve

shall restrici, our attention to

set-valuecl Íunctic¡ns wil,h r'àluesr

in

l,he subcone 6Goatu(El of Øoatu(E),

i! ihis

casetheconl,inuous set-valued

functions

coincide

rvith the

set-valuecl functions

rvhich

are continuous

rvith

respect

to the

symtnetric topologv

(cf., e.g.,

11,

Corollaly 1, p.

6?_l)

;

consequently, the

notation G('Yr66onu1l)) is

consistentl,vith

that

one rLsed by

Keimel

ancl

Roth and. a

set-valued

function f :X --

-->G6onz,(11)is continuous

if

ancl

onlyif, foreacn'

V

eE

anclcne

X,

there

cxists a

neighborhood rY

of

øo suoh

that

(1.6) l@) -

l@o)

I Y, l@ù . f@) t v

for

each

neN.

We

conclucte

this

Section

by lecalling thab in Ø(X, G€onu(El)

is definecl

the

follorving order

relation

(1.?) l<go,f(r)c:g(n)To,-eact-ue'Y

for

eaclr set-r'alued. functions

l,g X:,

--+ îîØonu(Ð).

We shall

also use

the

nol,al,ion

.l <

g

+

V

(1,

g

e(E(X,

V6o*u(E))

and V

eE) to

indicate

l@) -

g(n)

-l V lor

each u

eX.

2. tIp¡rroxirnation q¡f eontiruous sct-vniuetl funetions. fn this

Section rve

fix a

compact

llausdorff

topological space

X and a

Fréchet space

Il

1 we shall consider a sutrcon e Ø of

6(X,

G'€ ,rou(Ð)) cont'uining the single-valuerl

functions (i.e. {p} eç for

eah

9eØ((X, E))anct 'we

shall stucly

the

convergence

of

equicontinuous nets

of

monotone

linear

opera- {,ors

from 6 it G(X,

GGonø{E)).

First

of all, rve

recall

Lhat an

operator I

-,

Ø(Xr(66oaau(Ð)) fuorn

â, suboone G of

G(X,€6r,nu(E)linØ(X,Ø(€ontt(D))

is callerl linear

if r(f + s) : !I(l) I r(s), ?(À/) :

À?(/)

fcrr cach set-valuetl

functions l, I eG and

À

>

0.

(4)

:20 lrzfichele Carnpiti

¡tpro_rinratiotr of corr[ìnrrous s¡i-i,¿lr¡ccl lil nctiotrs TI 2l

IL is

¿leal

that

an

uniformly

conl,inuous op€ìÌ,atot,is

both

continuous

rì,1)r[ rnonoto]re

;

as a consequence of Lremma 2.11 a,lso

the

conyerse is true.

Cottc't,r,.\rr.r 2.2.

Let (l

bc

a s,tr

on,e

of ß(X, G$ooru)(Il))

cojttu,inirlfl

't,l¿e

r:on,sttntt,

set- I,

1ø\X ,'Ø(i ..tn,(

1tr)\'\

,â1X

,

ØØ o r,,L

(tt))

bc.t1

lincu,t

,,1Сi

iit ilset/.

'

Ihen,,

tlte

are'erly,i.ualent:'

a) T is

tr,rt,í.formlt1 con,tinu,ou,s..

b) ? fs

continu,ous and ,nt,onotoirc.

"i f

,l,s contLnnrous

at 0

ctnd, n¿otltotone.

.

.. r?roof

. ilhe implications a) > ìr) anil b) ='c)

¿ùre

trivial ancl

the

iurplicrbion

c) =+

a)

f'ollorvs fr,om' fremma

2.1.

H!.

ing Coroliaty

2.2,

il G

conlains bhe

tucly of the

con-r,ergenoe

of nets

of

from

(6

in 6(,1,

(6oan,(IX))

is

equi- ce

of nets of unilolnrly

cclnl,inuous

Iìr

1,he_ secluol,

rve-shall

considel subcones

.d of G(x,

.6Øotru(Il)) co-ntaining

the

single-valued

functions

ancl monotone contiriuous lineai trpclator:s 1' : (€ -n

6(x, Glîo^u(E))

satisfying

the folloriing

concutiorÌs:

(2.2) for

cach cg

eG(X,Il),

,I({91¡) is single-valuecl;

(2'3) for

each

f eG

antl

n €X?

r(l)

@)

-,.f,JrTQeÌ)

(ø).

|"v (1.1)' the identity

gperator satisfies concritions

(2.2)

uotd (z.s).

In the

second

palt

we shall see

that, in

1,he case

lt - R, moo"t"íe

continuous

linear

operators

ftom G(x, Rj in itself

generate

ii a

nal,ur,al

lYay monotone continuo[s linear 'opô"*tors

fuon"G(X,

,ArAoorrlþ¡l in

it;se

,lefi i'ä.hï :,'å!i,iiîå"1i,îî"î1Jïxiîifå.i*

fol

lls¡rilrrroN

2.5.

Let 6

ba ct

sub

cctntøini,ng the

single-auluecl

Jt nd,l

ntonotone con_

Iinuotts

l,iu,eor

sati

lf II

,is

ct, ,6, t:kin

set

in,

Ç

if (€(X,

,,fo, eaclr equicontinuous nct

(r,)f.¡

o.f nronotonelinear operators J'r.om

6

in,

6Gztzu(11))

suclt iltat ilte

n,et

(T,(lt))Í.t

"orrr"rgri to

T(h,)

for

eaclt,

h

eÍIr"ue

aL.so ltatse.tltø.í the rtet

(T,(Jl,<.¡

co??,Dlt.!cs

to I([) ,for

eaen¡

f eß.

-. -rl 'r is tltc

iden.ti'ty.^operurorr""a"ìt-T(orouhír, set

ir"rh"*ni ir'{¿*piir

cal[.ed, u,

](ot'otl;ùt

sc[ 'i,t¿ Ø.

As

observed

in l4l, if H

conLajns

the

constant sei,-valrred funotions, we carn

omit the

ecluicontinuity of the nel,

(2,)å, of

continuous monol,one

linear

operators

in Definition'2.8.

We

have

the

following

rnain

theorem,.

ruú ilcn

't'ot,,(D

'J'r;.1;:,f;:rr,l;r,i

ct,

,uì,ono

linctí,t oqrrr*iiì,

'tt,¡¡ .the,f

otknuing contl,í,t,iott

(2'4)

J',ot' eucrt .[

e'6,

no e

x

cøtcr

)/

e

E,

th,crc erists rt, e

rr

sttcr¡, ,¿ut

/ <

/¿

antl

rl,(tt) (øo)

.

,11(f)(no¡

¡

¡r,

l,l¿ut,

I{

,is a, ,!t_I(o¡,oul;in, set ,in, G.

IiLroo'f

'

LeL

(!t,)i'.¡

be an ecluicontin'ons

..ct

of

rnonot'.e

rjnea,r, ope'a-

l?iä.,ï1î '!ìiïì',,iÌt¿,^K;:Ií?;

ancr

s'ppose ilrar nie ner (,r,,(tt))Íet

con_

,'," ,ri:','iìf,:-i',iìit:i:T; lî*:'"

r;h¿t tr'¡e ncr,

(",(/))[¡

cor]\¡crs(,s

to

,1,(f),

3Il, ::¿i,ir

,sjuste v¿rtuect.

r,et

V e

E anct

consicler Ifl e

E s'ctr tirat ,f

and cron

.(z +l)¡.

rL.v y:iií ir!åT\i,TÍff1"1,,Æ#fl

hood

N(r¡) ,Ãìji;;: irr(i.tt.t +

2t4, *.r,en_

.i.¡or

earit teI, I,is

monotone ancl

thercjlore I,(Í) < ,j,,(h); on

1,he

<¡LltcÌ li¿nd' tlre

nct

(^r',(tt'))Íçr

"ooìru.g-à* ,Lo .,r(tr,)¿o,r tLìï.orore 1,hcl0 cxisr,s o:(n) e

r

suctr

trrat, ior'òáäì',;;;ì'3'

o(ro),

(2.b)

?,(l¿)

<

T(tt)

+Iv,

"(/i) ç T,(h)1_

fir.

) 3tr1'.antl for. eaclt n e

N(ns), I,(fl

(n)

c

e. Ilet ZeE

ancl

consider WeE

Let

îcne Z : try (2.4), thercexilf,s

/r,,1{¡rrctr,tat {.ç

t_anrt T(tr)

(rn)

c.

'T(J)

@o)

1-

Tiz';

moieoíã.,r,rrusriäir]i w¡ is

cornpact andtherefo.c {,here

erist

!/t¡. . . ,y,, e

T(fl (n) such thaL

1,

each

i!' 1,'.

.'.,;,;|.1=,"ií,J,*

"i_' _:I..,;"?

i: !,.

.

.,n

antl filrûhei,

,.I(f\(r^\c

vrrtue of thc continüíty"'of

(5)

Michele Catnpiti I Approxirnal ion of ctontitruon:;

REFI]I.ìENCIìS

St:t-y¿l¡p¡¡ lunctions I ¿,t

T({pr\),...,f({q"}), T(Í) tì,rd

af(/¿)

-\\¡e nra)¡

fiutl tttr open

neighbor'}roorLì rV(*,') oT *ro such l,hat,,fol e\¡ery n)

eN(nù,

(2.6) 2(l

)

(n) - I(,f)lr) +

2IV 1

rU)

@)

c

co

(Ir({erl'.

. .,Í({ç,,1 Xrx) J,-2['i¿.

At

1,his poinl,, lx'e obscìr've

that

{cp¿} is single-r.alucti

fol

ea,ch ¿

:

1

t.

. . t tt antt therefole

the net (f',([çù)),1,

conrrerges

lo

"([9¡]); moleovet',

sirlt:e h,

eII

ther

net

('1',(lt,))Íer cotl\¡et'ges

to

ft'; hence,

thele exists

ø.(a-;o)

e1

.qtlclh

tlrat, for

t¡a,ch ¿

eI, t)

ø"(ero)

ilntl j:7r...ttìt

(2.1) f,([ç,]¡ ç

?({<ai})

+ lt'', ?([q,]) < fi,({qil)-f ll/'

?,(/,) <'I(h)+W) fØ) <

",(l¿)+

l,l'.

Ìly

(2.6) an<ì (2.?) u'o ob1,airt, Tol eaoh ,' e

f ,

¿

2

a(';c,),

r

e,\r(øo) a'rtil .i

:

L,...trL¡

?({q,}) (r) c: I},([pi])

(¿r)

* fl'' c

'1',(.f) (n\

)- ll'

ilncl hcnce, sinco

?,(/)

(e:)

is

cottvex,

Q.9)

T U) (ar)

c ge(1.r(tp,ì,...,f'({ç,ì)

(,r)

* 2II; c

"'(.f) (r) -l 3'l'l'c:

'x,(h (r) ï lr

;

on the other

h¿r,ncl

(2.10) rl|)

@) c:'t-"(h ) (ø)

-

't'(h) (n)

-l- Ìl¡ c

'x(,1') @)

+

3Ì4/ c. T(l')(')i)

+v

-

Är'grring on

the

compastness of ,U a,s

ilr the first

case,

þ¡'

(2.9) a,nct

(2.10)'n'e tleduce

tlre

exislence

of ø.e1

such i,hal,,

lot'each

'. e

l,.t )

u-,

ftl) < f,(J) + lt, I,(,Ð.<

t'("f)

T'- zlrtd

tìris

c-omp1-ei,es-the

plooi.

In tllb

palticular; case u.here

the tipelrtor' ?

is

the

identitJ¡ opet'¿tto,'r

n'e

obtain the

liollorving Corollar'¡'.

Oor¡oi,t,¡l¿l: 2.6. f'et

X

be u' cornpnr:t ÍIctustl,or.f,f I'o1tolol¡ical s'pcr,ccr.Il-o

Frócltet sqtctce cund

G

a, stiltcott,e

of G(X,

(6Gottu(Il)) c:ottttrí,nin'gl'he si,n,gle- -ualu,erl Ju,rt ctions.

If" II is

u subset

oîG

sutisJyitt'(t tl¿e Jollotui'ng cott'd'il,ion

Q.E) for

eaclr' J e '6,

s;teX

a,t¡d

lr

e1{3-^, tltere enists lt

eII

suclt thttt

I < l,

ctnd tt'(n)

c f(r) ¡

V

'

tl¿en,

If is

u, K,oroul¡:in set in, ((,1

.

m

lìnlu.r,r

2.?

Ilntler

tlre hypotheses of

theolem

2.4,

if the suÌ;conc'l

also coltl;ains

the

constant set-valuetl l'unctions, conclitjorl (2.4) ma,l' 'l¡t¿

replacetl

by tÌre

follorving

(2.L2) fol

each .:[

e'€,

aoë,X.arLcl

I/

e

E,

there cxists /¿e

]f

such

that

I <

h

f -ll and 'I(tt) (nfi - 'I(l)

(#o)

+

Y.

'l'he proof is similar to

insteacl of the monotonioity o is applied

to

an elernent Û e 6(ú't

nu(fr))

ancl

if / <

g

+

L/, bhe

lVlor,eover,, ii' rve consirler, 1,lre

11 r'oplacerl

by thc

following.

(3.1-3)

fol

cach

Í e16, noeX-attl

V

/ <

/¿

_f Z

ancl tt(nn) c. l@o)

+

V.

(p

-L. e 'lB

:

ir,(ø), rvhile, for each 'llS"c/(øo) -F

(y(r)

-l-

") .

[B:

å(n,);

e ' Its,

thc ploof

isl

cornplete. g

i a I i nc I u ç iott s, Glunrìlchlcn der a Lhcnì atisclÌcn for II,29__35.scl_ualuecl Ilausclorff conlinuotts furtcíions t lype approximdlion lltcorcü.r for sel_ucrluccl ftutc_

8t 9-.s2Íi.

a t u.l. ct p p r o.t i t nallon, p r.c¡r r.in t .I,ecìruischc I lrr chs.

- -s9.

n. t,Ia[lr., 6J (1956), 2, 961 _982.

ualued Iuttt:liotrs, J. Approx. .I,hcoly, ¿0 (1g?g)

Iìecoivcd 1.IX.19gO

Diparlintetúo cli Matentatica Unioetsitì! degli Sttttli tli ßari

if tauetsa 200 Via Re Dauid, ,l

't 012! BARI (17'ALy )

i

l

I

liltr tl lrl

Ì

ii

ll

l

Referințe

DOCUMENTE SIMILARE

Several generalizations of the classical notions of convexity and quasicon- vexity of real-valued functions have been given for vector-valued functions and even for set-valued maps

Lipschitz duals, meaning spaces of Lipschitz functions on a metric linear space, were used to study best approximation problems in such spaces (see [10]).. A good account on

singular integrals of Picard type which approximate the continuous functions defined on compact intervals... Th, Anghelulá, une remarque sur I'integrale de poisson,

Devore, The Approximation of Continuous Functions by Positive Linear operators, Lecture Notes in Mathematics, 293, Springer Yerlag,7972'. Starrcu, Approxintation of

Devore, The Approximation of Continuous Functions by Positive Linear operators, Lecture Notes in Mathematics, 293, Springer Yerlag,7972'.. Starrcu, Approxintation of

Cliteria fol the contjìluity of convcx and concave functjons play ar irnpoltant r'olc in fructional :ùnâlysis ¿ùlìcl optimization theoly, no nlatter u']ret,her

Iulrodrrctiull&#34; '-l.he l.ell-lcno\\¡rì resulbs on the Korovkin ¿ìpploxima- 1,ion o[ continlrous leal-valuec1 functions h¿1,'se playecl a crucial rôle jn the

category, tken there exist, for each family ff of real-aalued lower semicontinuous func- tions on X ah,ich is þointuise bounded. Singularities of Families of