Monday, April 9, 2007

Ipod Shareaza Connect

Minimization of convex functions. The

can give two answers, well known
on the minimization of convex functions: The first

. We recall the following formulation

(perhaps more practical):

Theorem.

Let (E, II) a reflexive Banach space.

A is a subset of non-empty, convex and firm E;

f: A ----]- \\ infty, \\ infty] is a convex function

, semi-continuous - inferieuremet, non-

-identity equal to \\ infty and such that:

lim f (x) = \\ infty <= k max(rho(x-y),rho(x-Tx),rho(y-Ty)) sur DxD (*)
IXI -> \\ infty
As a simple example f (x) = Ix-aI. (A in E)

For proof of this result, see

H. Brezis (References 02/02/2007).

deuxieme.Le The following result generalizes a

resulat well known in the real right

:

Theorem.

Let E be a Banach space is a convex quelconque.f

IR.On in E implies that there exists a z in E such that

d_ {+} f (z) = 0.

(d_ {+} f (z). X = lim [f (z + hx)-f (z)] / h,
h -> 0 ^ {+}

with z, x E h IR)


Then f has a minimum at z

for information on d_ {+} f (z), see

RHMartin (References 02/02/2007).

Then, we easily show the inequality

f (x)> = f (z) + d_ {+} f (z). (Xz)

Finally, note that if f is strictly convex, then , f admits a unique minimum

.


0 comments:

Post a Comment