Вы находитесь на странице: 1из 6

519.6:004.

93
.. , ..., . , . ,
.. , ..., . , . ,
.. , ..., . , .



.

. .: 21 .
: , , .

.
() ,

,
t [1 3]

J (c, t ) M x {Q(x, c, t )} ,

(1)

J (c, t ) ; Q(x, c, t )

c (c1 , , c N ) ; x ( x1 , , xM ) [4, 5].


.

,
[6 11],

c[n 1] c[n] [n]g (c[n]),

(2)

c[n] n ; g (c[n])
; [n] .
-
,
.

[6 13].
,
,
, [7].
197


[14, 15].

, [5].

.



.
.
[4, 5, 16, 17]
s

c[n] c[n 1] [n]

m 1

m 1

m [n] c Q( x[n], c[n 1], a[n m])

m [n] c Q( x[n], c[n 1], a[n m])

(3)

Q(x, c)

ci , i 1, , N ; m [n], m 1, , s [n]
- ().

, N . ( )
,
1
( i)
,
x
[18], ,
.
, ,
1. ,
, ,
,
1
( ,
).
x
,
(1), c* [n]

198

c *[n] U (n, n 1,, n k , c *[n 1],, c *[n k ]),

(4)

U , "",

l 0, , k c* [n l ] , .
(1),
, [19], (5) (6) (7),
n (8) [4, 20], (4) (8),
(9). a 0 , [n]
.

c *[n] a1n a0 ;
c * [n] (1

(5)

a
1
)c * [n 1] 0 ;
n 1
n 1

(6)

a0
;
n 1

(7)

c*[n] U (n, n 1, c*[n 1])

c *[n] U (n, n 1, c *[n 1]) ;

(8)

~
c[n] U (n, n 1, c[n 1]) [n] c Q( x[n],U (n, n 1, c[n 1]), n) .

(9)

, (9)
c* [n] A
n. , n (9)
2

, 2 [n] M c[n] c* [n] , ,


U (10) , U
(,
[10])

c * (t ) U (t , t 0 , c * (t 0 )) .
[4],

(10)

bn , 0 bn 1,

n1

lim Wn W ,

lim bk

n k 1

(1 b )W
l

l k 1

199

bn , W n

(1 bl ) 1 ,

l k 1

, [n] [n] ,

~
c[n] U (n, n 1, c[n 1]) [n] c Q( x[n],U (n, n 1, c[n 1], n)) .

(11)

. :
1. x[n] .

~
~
2. M { c Q( x[n], c, n) M c Q( x[n], c, n)

/ c} 12 [n] 22 [n] c c *[n] ,

12 [n] 22 [n] .

~
3. M c Q( x[n], c, n)

4. [n]

S 2 [ n] c c * [ n] .

U (n , n 1, c) U (n , n 1, c* [n])

1 [n] .

c c* [ n]

2
~
5. (c c* [n], M c Q( x[n], c, n)) s[n] c c* [n] , s[n] 0 .

6. 0 b[n] 1 (1 [n]) 2 2[n]s[n](1 [n]) 2 [n](1 2 [n])( S 2 [n] 22 [n]) 1 .

7. b[n] . 8. W [n]
n 1

9. 2 [0] M c[0] c* [0]

2 [n]12 [n]
0 n .
b[n]
2

; M c[n] c* [n]

0 n .

. (10) (11) [4, 12]

c[n] c *[n]

U (n, n 1, c[n 1]) U (n, n 1, c *[n 1])

2
~
2 [n] c Q( x[n],U (n, n 1, c[n 1], n) 2[n]( U (n, n 1,

~
c[n 1]) U (n, n 1, c *[n 1]), c Q( x[n],U (n, n 1, c[n 1], n)).

, 1 5 (12) :
200

(12)

2 [n] ((1 [n]) 2 2[n]s[n](1 [n]) 2 [n](1


[n]) 2 ( S 2 [n] 22 [n])) 2 [n 1] 2 [n]12 [n];
b[n] 1 (1 [n]) 2 2[n]s[n](1 [n])
2 [n](1 [n]) 2 ( S 2 [n] 22 [n]));
W [k ]

2 [k ]12 [k ]
.
b[k ]

(13)

(14)

(15)

(13) [4] 6 9 [4],


. c[n] c* [n]
1

2 [n]12 [n] .

n 1

[n] [n] , , , 6 8

[n] ,

min [n]
u n , u , , 0 u 1
[ n]
1, , N . [4, 21].
.
,
[1].
: 1. ..
. / .. , .. , .. // . 2009. . 72. . 5561. 2. ..
- / .. , ..
// . 2009. 6 (40). . 123127. 3. ..
/ ..
// . 1976. 11. . 91101. 4. ..
/ .. , .. ,
.. // . 1970. 5. . 920. 5. ..
-
/ .. , .. // . . . - . . -
... 2008. . 12. . 5660. 6. ..

/ .. //
. 2009. . 5. . 2439. 7. ..
. .: . 1976. 488 . 8. ..
/ .. ,

201

.. // . 1979. 5. . 8998. 9. ..

/ .. // . 1969. 2. . 7383.
10. ..
/ .. // . 1975. 3. . 7379. 11. ..
/ .. // . 1974. 3.
. 131133. 12. ..
/ .. . .: , 1968. 400 . 13. ..
/ .. , .. //
. 1990. . 26. . 4553. 14. ..
/ .. . .: . 1983. 384 . 15. ..
// .. . : . " -". 2007. 87 .
16. ..
- / .. , ..
// . . . "., . . ". 2010. . 12 (165).
. 162168. 17. ..
- / .. // . . . -. 2007.
. 1 (27). . 207213. 18. Krylov V.N. Contour images segmentation in space of wavelet
transform with the use of lifting / V.N. Krylov, M.V. Polyakova // Optical-electronic informativelypower technologies. 2007. 2 (12). P. 4858. 19. ..

/ .. , .. , .. . .: . ., 1976. 124 . 20. Dupac V.


A dynamic stochastic approximation method / V. Dupac // The Ann. of Math. Stat.1965. V. 36.
6. P. 16951702. 21. .. . / ..
// . 1965. . 10. .2. . 2731.
519.6:004.93

/ .., .., .. // "".
: . : "". 2011. 36. . 197 202.

,

. .: 21 .
: , , .
UDC 519.6:004.93
Multi-starting sub-gradient optimization method under no stationary conditions
/ Shcherbakova G. Yu., Krylov V.N., Logvinov O.V. // Herald of the National Technical
University "KhPI". Subject issue: Information Science and Modelling. Kharkov: NTU "KhPI".
2011. 36. P. 197 202.
Multi-start sub-gradient optimization method under no stationary conditions is designed and
investigated. This method allows noise stability raising, error, local extreme sensitiveness reducing
in the procedures of the technical diagnostic under no stationary conditions. Refs.: 21 titles.
Key words: sub-gradient method, optimization, no stationary conditions.
16.06.2011

202