Discrepancy principle for DSM. II

Abstract: Let

A y = f

,

A

is a linear operator in a Hilbert space

H

,

y p e r p N (A) := u : A u = 0

,

R (A) := h : h = A u, u i n D (A)

is not closed,

| f_{d} e l t a - f | l e q d e l t a

. Given

f_{d} e l t a

, one wants to construct

u_{d} e l t a

such that

l i m_{d e l t a o 0} | u_{d} e l t a - y | = 0

. A version of the DSM (dynamical systems method) for finding

u_{d} e l t a

consists of solving the problem �ee dotu_delta(t)=-u_delta(t)+T^{-1}_{a(t)} A^ast f_delta, quad u(0)=u_0, eqno{(ast)}eee where

T := A^{a} s t A

,

T_{a} := T + a I

, and

a = a (t) > 0

,

a (t) s e a r r o w 0

as

t o i n f t y

is suitably chosen. It is proved that

u_{d} e l t a := u_{d} e l t a (t_{d} e l t a)

has the property

l i m_{d e l t a o 0} | u_{d} e l t a - y | = 0

. Here the stopping time

t_{d} e l t a

is defined by the discrepancy principle: �ee eqno{(astast)}eee

c i n (1, 2)

is a constant. Equation

(a s t)

defines

t_{d} e l t a

uniquely and

l i m_{d e l t a o 0} t_{d} e l t a = i n f t y

. Another version of the discrepancy principle is also proved in this paper.

Recommendations

Cites Work

Cited In (4)

This page was built for publication: Discrepancy principle for DSM. II