The Computation of Fixed Points and Applications

Fixed-point algorithms have diverse applications in economics, optimization, game theory and the numerical solution of boundary-value problems. Since Scarf's pioneering work [56,57] on obtaining approximate fixed points of continuous mappings, a great dea

  • PDF / 7,424,297 Bytes
  • 138 Pages / 397 x 612 pts Page_size
  • 89 Downloads / 239 Views

DOWNLOAD

REPORT


Managing Editors: M. Beckmann and H. P. KUnzi

Mathematical Economics

124 Michael J. Todd

The Computation of Fixed Points and Applications

Springer-Verlag Berlin Heidelberg GmbH

Editorial Board H. Albach . A. V. 8alakrishnan . M. 8eckmann (Managing Editor) P. Dhrymes . J. Green . W. Hildenbrand . W. Krelle H. P. Kunzi (Managing Editor) . K. Ritter . R.Sato • H. Schelbert P. Schonfeld Managing Editors Prof. Dr. M. 8eckmann 8rown University Providence, RI 02912/USA

Prof. Dr. H. P. KOnzi Universitat ZOrich 8090 ZOrich/Schweiz

Author Michael J. Todd School of Operations Research and Industrial Engineering Cornell University Ithaca, New York 14~53/USA

Llbrary or eoa,reu Cat.loela, la Publicatieu Data

Todd, Michae1 J 1947The ccmputation of fixe

O}

be continuous. and

-,

C.

~

= C..

f. (x) > x.

then

~

~

Ci ~ Sn, i E I.

J

= NO

=

cj

~

UjEJ Cj

f.

is continuous and

we have

x* E c~, x''; i'

0 Note:

sat-

c~,

i E No'

'

Since for x E

rt

i E NO'

s~,

x ~ C~,

and condition (ii) is verified.

If the K-K-M lemma holds, we can deduce the existence of f

i E NO'

=

Also, by definition, if

x E UjEJ

~ I,

1

T v x

Sn ~ UiEN C: c U. N C.. o ~ - ~E 0 ~

Hence

Ci ,

lies in none of the

Then

condition (i) is verified. Thus for

If

x E Sn

for

a contradiction.

let

i E NO

We now show that the

~

isfy the hypotheses of the K-K-M lemma.

For

x*

f. (x;,)

i E No' and so

~

Now

1

x*

Since

C ••

1 =

is a fixed point of

Scarf [58] has proved a lemma similar to 3.1 in which condition (ii) is

replaced by: (ii) ,

For all

S~ c C ••

i E No'

1

-

1

One can easily verify that this version also implies Brouwer's theorem for

Sn.

Freidenfelds [22] has shown that the two versions are equivalent. 3.3

Simplices and Triangulations.

3.1 by exhibiting, for any

s

>

0,

one another, with one point in each Sn

into many small simplices.

As we discussed in 2.6(c), we shall prove

an (n+l)-tuple of points of C •• ~

Sn

within

of

We produce these (n+l)-tuples by dividing

The concepts of simplex and triangulation will be

B

introduced informally now; we postpone rigorous definitions until Chapter III. A closed j-simplex is the convex hull of finely independent) in

j+l

points in general position (af-

the points are called the vertices of the simplex.

An

open j-simplex is the relative interior of a closed j-simplex.

Closed n-simplex

n

Note:

Sn

o

(point)

1

(line segment)

2

(triangle)

3

(tetrahedron)

is a closed n-simplex.

A face of a simplex Thus

s~

Open n-simplex

a is a simplex all of whose vertices are vertices of a.

is a closed face of

Sn

for each

i E NO'

Two simplices are incident if one is a face of the other.

Two j-simplices are

adjacent if they share a (j-l)-simplex as a face. A triangulation

G of

Sn

is a finite collection of open n-simplices such

that the open n-simplices, together with all their open faces, form a partition of Sn,

i.e.,

Sn

is their disjoint union.

As we shall see in Chapter III, this rather

non-intuitive definition, when couched in