194

Click here to load reader

Дискретная оптимизация и исследование операций: Материалы российской конференции (Владивосток, 7-14

  • Upload
    dinhtu

  • View
    249

  • Download
    8

Embed Size (px)

Citation preview

  • . .-. . . . ()

    . .-. . . . ()

    . . . . . ()

    . .-. . . . ()

    . .-. . . . ()

    . .-. . . . ()

    . .-. . . . ()

    . . ()

    . . ()

    . . . . . ()

    . .-. . . . ()

    . .-. . . . ()

    . .-. . . . ()

    . . . ()

    . .-. . . . ()

    . .-. . . . ()

    . .-. . . . ()

    . .-. . . . ()

    . .-. . . . ()

    ..-.. . . ()

  • . ..

    , 7 14 2007

    2007

  • 22.1 482

    -: (, 7 14 2007). : - , 2007. 192 .

    ISBN 978-5-86134-134-9.

    , (DOOR-07).

    : . . , . . , . . , . . , . . , . . . . . , . . . . . .

    :

    ( 07-01-06052)

    20070382011602100000

    )( .

    ISBN 978-5-86134-134-9 . . . , 2007

  • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

    . . . . . . . . . . . . . . . . 94

    . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

    115

    . . . . . . . . . . 142

    . . . . . . . . . 149

    . . . . . . . . . . . . 157

    . . . . . . . . 173

    . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

  • . .

    -

    -

    . ,

    .

    -

    -

    , ,

    -

    .

    -

    , -

    () . -

    .

    -

    ( ), -

    , .

    -

    [1, 2].

    x

    = (x1, , xn),

    :

    x

    = argmaxP{X(x

    , t) Dx

    ,t [0, T ]} (1) X(x

    , t) ; Dx

    -

    ; T . D

    x

    , , -

    , :

    aj y(x)j bj, j = 1,m y = {yj}mj=1 , yj = Fj(x1, , xn), Fj() , . , -

    , -

    -

    .

    -

    .

    (-) (-

    -

    ) . N ,

    4

  • , -

    .

    -

    .

    -

    , , .

    -

    -

    , , -

    . ,

    , -

    , . ,

    ,

    ,

    .

    -

    . -

    ,

    , -

    . (1)

    - (master-slave) -

    .

    -

    ( ),

    -

    .

    -

    [3, 4]. -

    .

    , -

    .

    1. .. . -

    . . : . 1992.

    2. .. . -

    // , 4, 2006. . 38.

    3. .. , .. .

    . // , 4, 2005. . 1926.

    4. .. , .. . -

    . // , 4, 2003. . 1115.

    ,

    , . , 5, ,

    690041, , (8-4232) 31-02-02, E-mail:[email protected]

    5

  • 1

    . .

    1

    , -

    = - - p = p , w = w -

    , f(w) Min{, f(w) | g(w) p, w }, (1)

    f(w) , 0, 0, (2)g(w) p, p p 0, p 0. (3) f(w), g(w)- , - , f(w) Rm, g(w) Rm1 , w Rn, Rm+ , p Rm1+ . (1) p 0 - ( ) ,

    f(w) D(p) = {w Rn | g(w) p, w }, [1]

    f(wy) ParetoMin{f(w) | w D(p)}. (4) f(wp) , - () K(f(wp)) f(wp), w

    p D(p), .. K(f(wp)) = {f Rm | f f(w)} - f(D(p)) f(wp), .. K(f(w

    p))

    f(D(p)) = f(wp). f(wp) - . , (1)-(3) ,

    wp , - f(wp) -

    ( p 0). , 0 p 0 , - , -

    w D(p), f (w) Rm+ p

    Rm1+ . - , 0, p 0 f (, p), p(, p). - ,

    (1)-(3).

    1

    ( 05-01-00242)

    ( -2240.2006.1)

    6

  • 2 -

    ,

    - ,

    - .

    , (1)-(3) p 0 - (1) .

    , f(w) Min{, f(w) | g(w) p, w }, (5)f(w) , 0, 0. (6) D = {w | g(w) p, w } (5) , f f(D) = {f = f(w), w D} f(w), w D - . - f(D) f , " " " ": fi = fi(w

    ) =min{fi(w) | w D}, i = 1, ...,m. wi , w f(w) " " f . , f 0, (5) , .. i 6= 0 , f < 0, . f , i, i = 1, 2, ...m , - . ,

    .

    . (5),(6) ,

    fi(x), i = 1, 2, ...,m fi(w)

    fi(w) > 0 w (7) , i = 1. (5) -

    f(w) D. ( ) . , -

    f(w) -, , f, f = f(w) w D f(w), , f , f f = f(w), w D, .. , f f 0. , -

    -

    .

    (6) , -

    f(w) , = 0, (8)

    f(w) 0. (9)

    7

  • , 6= 0, (8)

    = f(w). (10)

    (8) = 0, (9) f(w) 0. , f(w) ,.. fi(w) 0, i = 1, 2, ...,m, i 6= 0, i = 1, 2, ...,m. , . , (7) (6)

    (10).

    (5) -

    ,

    ( ) (

    ). w, f(w) ( ). () -. , -

    (5),(6) ()

    (10).

    .

    (6) ,

    ( ) -

    argmax{, f(w) 12 | 0}. (11)

    (5),(6)

    , f(w) Min{, f(w) | g(w) p, w }, (12)

    = argmax{12| f(w)|2 | 0}. (13) ,

    .

    3

    , 0 (1)-(3) - .

    , f(w) = (w), (1)-(3)

    w = Argmin{(w) | g(w) p, w }, (14)

    g(w) p, p p 0, p 0. (15)

    8

  • (w) : Rn R, g(w) : Rn Rm1 - , Rn- , p Rm1+ . D(p) = {w | g(w) p, w } - p 0. p 0 -. -

    ( )

    - () p 0

    f(p) = min{(w) | g(w) p, w }, (16)

    f(p), p 0 - [2]. :

    1. f(p) - , - .

    2. f(p) - , f(p) - - p Rm1+ . p , , , -.

    3. f(p) p (14) p - .

    (14) L(v, p) = (w) +p, g(w) p, w , p Rm1+ ( p ). , p 0 (14) ( ) , , p - w, p .

    L(w, p) = (w) + p, g(w) p w , p Rm1+ . (17)

    f(p)

    f(p+4p) f(p),4p 0 p 0,4p 0.

    f(p) - , .. p : Rm1+ Rm1+ . - . ,

    -

    p 0, f(p) : Rm1+ Rm1+ . p 0 - (14),(15).

    . ,

    (14), (15) -

    , ,

    p 0. -

    w ,

    9

  • f(p) (14). , (- p) , () , , .. -

    ( ) ,

    -

    . -

    . - , ,

    -

    , ,

    . , f(p), - p r(p), p. , p 0, f(p) = r(p). , () () .

    , -

    . ( p) ""f(p) r(p). -

    f(p) r(p), (), .. [3]. r(p) r(p), - , .. f(p) r(p) f(p). (14),(15). L(w, p, y(s)) M(w, p, p)

    M(w, p, p) = (w) + p, g(w) (1/2)p w , p Rm+ . (18) w ( -) p Rm1+ . , , . w p Rm1+ .

    (w) + p, g(w) (1/2)p (w) + p, g(w) (1/2)p (w) + p, g(w) (1/2)p w , p Rm1+ . (19)

    p Argmax{p, g(w) (1/2)p | p 0}. (20) , (19) (20) - p, w -

    w Argmin{(w) | g(w) p, w }. (21)

    4

    (1)-(3), -

    . (11) (20) (1)-(3)

    10

  • , f(w) Min{, f(w) | g(w) p, w }, (22) argmax{, f(w) (1/2) | 0}, (23)p argmax{p, g(w) (1/2)p | p 0}. (24) (22)-(24) -

    [4]:

    ()

    wn = argmin{12|w wn|2 + (n, f(w) n+ pn, g(w) pn) | w }, (25)

    ()

    n+1 = argmin{12| n|2 (, f(wn) (1/2) | 0}, (26)

    pn+1 = argmin{12|p pn|2 (p, g(wn) (1/2)p | p 0}. (27)

    wn+1 = argmin{12|w wn|2 + (n+1, f(w) n+ pn+1, g(w) pn) | w }. (28)

    1 (1)-(3) ,

    f(w),g(w) , - , wn, n, pn (25)-(28) > 0, , .. wn, n, pn w, , p n w0, 0, p0. , ,

    .

    1. .. , .. , .. . -

    . .: , 1986.

    2. S. Zlobec. Stable Parametric Programming. Dordrecht -London.: Kluwer Academic

    Publishers. 2001.

    3. .. . .

    . .5. .: , 2005. . 148156.

    4. .. . -

    // . .

    . . 1995. . 35, 5. . 688704.

    , . .. -

    , . 40, , 119991, . tel: (7-495) 135-

    42-50, fax: (7-495) 135-61-59. E-mail: [email protected]

    11

  • ,

    ,

    . .

    -

    . -

    -

    () ( ) (. [1-3]).

    ( 1904.) -, ( )

    -. , , -

    , (1874.) [2,3,6].

    -

    . Ax 0, .. A1. :

    Ax 0 Ax y = 0, y 0 x = A1y =nj=1

    ajyj, y 0,

    .. Ax 0 {aj j 1, n} A1. , {x =nj

    ajyj, yj 0} = C, C - Ax 0 Ax 0. , x 6= 0. m n Ax 0 (. [4]).

    .

    X Y Rm Rn ; A = (aij)mn X. v = max

    XminY(x,Ay)

    X ( ). v = minY

    maxX

    (x,Ay) Y

    (A). v ( -

    ). , A Y . Y : w = max

    YminX

    (x,Ay) -

    Y , w = minX

    maxY

    (x,Ay) X. C

    w A X X., {v, v} {w,w} - max, min . v+ = max{c, x | Ax b} - v+ = max

    xminu0

    {(xu) = (c, x) (u,Ax b)}. - (.. )

    v = min{(c, x) | Ax b} . , v+ x 0 max

    x0, -

    v x 0. (x 0) : v+ b;

    12

  • , b. : max{(b, u) | uA = c, u 0} min{(b, u) | uA = c, u 0}. , , : max min . . 1. v v+ . (xu) Ax b. . Uk ={u Rm |

    mj

    uj = k, uj 0} (k > 0) : v+mink

    maxx

    minUk

    (x, u).

    , , ,

    v = mink>0

    max{(c, x) + kt | Ax+ tem b}.

    em = (1 . . . 1) Rm. v+ v = max

    k>0minx

    maxUk

    (x, u) = maxk>0

    min{(c, x) +ht | Ax+ tem b} .

    max{(c, x) | Ax em} = k min{(c, x) | Ax em} = k. f(k) ( max) v. v = min

    k>0f(k). v = max{(c, x) | Ax b}. 2. k k f(k) k k k, v , v = v = f(k) k =

    ni=1

    ui, (u1 . . . um)

    v. (ai,x) bi, i 1,m

    v+ = maxx

    mini{(ai, x) bi}, v = min

    xmaxi{(ai, x) bi}

    v+ = max{t | Ax b tem}, v = min{t | Ax b tem}.

    v+ v (v+ > 0) (v+ < 0) Ax b. Ax b.

    v+ = max{(b, u) | uA = 0, (u, e) = 1, u o}v = min{(b, u) | uA = 0, (u, e) = 1, u o} v v+. - [2,3]. , v+ v (x, u) = (u,Ax b) (ue) = 1 u 0 .

    13

  • cj (j 1, s) bj (j 1, k), .

    L : v = max

    min

    maxx

    {(sj

    jcj, x

    )Ax kj

    bj,

    }

    kj

    j = 1 =sj

    j, j 0, j 0. , L

    v = max

    {x0|Ax

    kj

    bj, j, x0 (cj, x) (j 1, s),kj

    j = 1, j 0}.

    L min

    max

    minx{ . . . } .

    ()

    c cj j 1, s Ax b:1) x (c1, x), . . . , (cs, x). - C(x) = ((ci, x)/(cj, x))ss - ( )

    . , , - . -

    Ax b.2) Ax = b, x 0 s

    x =s

    j=1

    xj xj 0 : (sj = 1 j 0).

    C : minj

    maxxj

    {sj(cj, xj)

    sAxj = b, xj 0

    }.

    : n = 1 {x1 = 1 x1 0} 1/c1, . . . , 1/cs . C

    minj

    maxxj

    {sj

    jxjcj

    sxj 1 xj 0

    }.

    : j = ci/scj -

    : (EA)x = c, x 0, A = (aij) 0, E nn, x, c 0 Rn. , c > 0 x 0 (aij i- j- ). [5]

    (, A > 0).

    14

  • hi(x) =nj=1

    aijxj i- xT = (x1, . . . , xn)

    (i 1, n);hi(x)/xi () i- x;maxi{hi(x)/xi} x;

    mini{hi(x)/xi} x;

    h = minx>0

    maxi{hi(x)/xi} - ;

    h = maxx>0

    mini{hi(x)/xi} . , x0 > 0 (Ax0 = 0x0, 0 < 0 < 1). h = 0 =h = hi(x

    0)/x0i , i 1, n. . ki = 1/x

    0i , i 1, n K ki > 0. A(K) = (aij) = KAK

    1; e =

    (e1, . . . , en) A = (aij); e= (e

    1, . . . , e

    n) ( ) ei = kie

    i.

    aij = aijki/kj .. A(K) e.

    A 1/x0i , - 0 < 1, .

    , .

    -

    1) (EA)x 0 2). (EA)x 0 (A-). - - 1) C = {y = (E AT )u, u 0}. (2) K = {y = (EA)1x, x 0}. 3. A K C. -, .. (E A)1 0, K Rn+, (E AT )u = yu 0 y Rn (AT -). : (E AT )u = (E A)1x u 0, x 0, x 0. (EA)(EAT )u = x u 0 x 0. A = A+ AT AAT . 4 ( A). A-, 1)x 0 (E A)y = x, y 0;2)(EA)1 = BTB > 0;3) A 0, A ;4)E A . B = E + AAT , A = A+ AT . 5. A , BxAx c, x 0 , .. c 0; B 0 ,A .. A . 1) ATx A1y = 0, x 0 y > 0 ;2)AATx = y, x 0, y > 0 ;3) - Ax 0

    Ax 0 .

    15

  • .

    3) .

    . ,

    AAT A + AT . aij = aijaji i- - i- j. A = (aij) = A AT . A . 6. A , A (E A) - .

    , , i- A i- A2, : (E A) A .

    , , -

    , (-

    ) ,

    . -

    , (

    -).

    ( 07-01-00399) -

    ( -5595.2006.1).

    1. .. . . . , 1998, 248 .

    2. . .. , 1959, 470 .

    3. .. . . , 1968, 468 c.

    4. . . .: , 1973, 470 .

    5. P. . .: , 1989, 656 .

    6. .. . . .: , 1982, 152 c.

    , ,

    . ., 16, , 620219, ,

    . (343)375-34-28, (343)374-25-81, E-mail: [email protected]

    16

  • . .

    ( )

    . -

    -

    , () ,

    . -

    ,

    .

    , , -

    -

    .

    ,

    , .

    ,

    .

    ,

    .

    :

    1. .

    2. , ,

    .

    3. ,

    ,

    .

    , , , -

    ,

    , -

    , -

    .

    -

    -

    -

    :

    max(xi),(xij)

    {iI

    fixi +jJ

    iI

    djxij jJ

    iI

    dj zij

    };

    iI

    xij 1, j J ;

    17

  • xi xij, i I; j J ;xi +

    lji

    xlj 1, i I; j J ;

    xi, xij {0, 1}, i I; j J ;(zi), (zij) :

    max(zi),(zij)

    {iI

    gizi +jJ

    iI

    djzij

    };

    iI

    zij 1, j J ;

    zi zij, i I; j J ;xi + zi +

    lji

    xlj 1, i I; j J ;

    zi, zij {0, 1}, i I; j J.

    dj j J ;fi i I -;

    gi i I -;

    j I, j J . i j l, j i l. i 4j l , i j l i = l.

    -

    .

    (0, 1) w = (wi)(i I) I0(w) = {i I|wi =0}. j J ij(w) i0 I0(w), i0 4j i, i I0(w). I0(w) = , ij(w) i0 I, i0

  • u :

    minu

    {g(u, y) =

    iI

    giui +jJ

    dj

    iji0(y)ui

    };

    ui {0, 1}, i I.

    (0, 1) y u. (0, 1) y, - u (0, 1) y u , f(y, u) f(y, u). , ,

    ( ) (0, 1)y, y u - (y, u) f(y, u).

    -

    (0, 1) y (0, 1) ys, s = 1, . . . , S, - -

    . ys us - (ys, us), s = 1, . . . , S, (y0, u0) f(y, u). y0 -.

    ys (0, 1) ws, u1, . . . , us1 . . -

    . , -

    , -

    .

    060100075.

    1. .. -

    . : , 2005.

    . . . . 4,

    , 630090, , . (383) 333-28-92, (383) 333-25-98,

    E-mail: [email protected]

    19

  • . .

    -

    , -

    .

    - -

    .

    1. , , -

    (Q,), Q - , - - - (. [1]). , , [1]

    , Q - d, = B, B - - (Q, d).

    (. -

    3 ). K- [2], [1,3,4].

    , (Q, d) - d, B - -. V v : B R, v() = 0. - [4,5] = (Q,B, v) , Q -, e Q, B - ., v(e) e. , , () v.

    , -

    v(Q) Q . , (v), - 1953 . -

    . , .

    ( ., , [4]) ,

    v = f , f - [0, 1], - - B., - C(v). - v(Q)): C(v) e B. . . [6] , ,

    , ,

    . C(v) .

    2. ,

    . e - B. H(e) -

    20

  • B- e H = eBH(e)., = {ei}m1 H v V v() = v({ei}m1 ) v,

    v() =

    (1)m||v(iei),

    = {1, . . . ,m}, = {ei}i, || - . v().

    1 [3]. vo = sup { |v()| H(Q)} v. , v V , vo

  • rV , o o , .. (o)- - o. -

    rV, (. [7]).

    4 [3]. v rV n, n + 1 : {ei}n+11 H v({ei}n+11 ) = 0. - rV n.

    5 [3]. rpV = n=1rV n. rpV () .

    6 [3]. , v rV n n, rV n1 (.. |v| |u| = 0 u rV n1). n rV (n).

    rV n, rV (n) rpV .

    1. rpV (-) rV (.. v rpV |v| o |u| u rpV ). 2. n 1 rV n rV (n) - () rV .

    2 , , v rV m 1 v(m) rV (m):

    v(m) = sup{u rV (m)+ | v+ o u} sup{w rV (m)+ | v o w};

    ( ) v rV+ v(m) rV (m)+ , rV

    (m)+ = (rV

    (m))+ - 1

    rV (m). , v rV v(m), m = 1, . . . , - .

    v rV, , v(m), v.

    7 [3]. v rV () -, : v =

    m=1 v(m), - o. rV raV.

    -

    , [3] -

    , ..

    1. ,

    , .

    1

    W V rW+ =W rV+.

    22

  • 1 ( ), -

    (

    o) . -

    .

    3. , n- . v rV n - v B[n] B, . e B e[n] = { e

    | | n}, , , | | . 8 [3]. Q[n], - {e[n] | e B}, B[n] ( n) B.

    e[n] H: = {e1, . . . , em} H

    [n] = { Q[n] mi=1ei, ei 6= , i = 1, . . . ,m}. B[n] .

    3 [1]. B[n] - [n] ( H). E B[n] - H(Q) (, E) , ( ):

    E =

    (E,)()[n].

    v - v B[n], -

    v(E) =

    (E,)v(),

    (E, ) E ( v 3). - v Q, [3], v - - v - B

    [n], - B[n] ( {f1, . . . , fm}[n] -, {fi}m1 H, fi F, i = 1, . . . ,m; . [3] ). [3], B[n] . -

    , d[n](, ) = min { , }, , - -

    , Q[n], , d[n] d Q[n], - Q[n] , , (Q[n], d[n]) - . , .

    4 [3]. B[n] - - (Q[n], d[n]).

    23

  • , Q Q(n) = { Q[n] | | = n} Q(n) B[n] \ B[n], n = 2, . . . , , n 2 B[n] B[n]. f - I(Q,B)

    B- , c - , 2 m n, , c(< x1 >) = x1 x1 R. fnc : Q[n] R

    fnc () = c(< f(t1), . . . , f(tm) >), = {t1, . . . , tm} Q[n]. c ( n), fc = f

    nc - c- f., , .

    9. f I(Q,B), v rV n(B), c n. , f (v, c)-, Icv(f) =

    fcdv =

    Q[n] f

    nc dv ;

    fcdv (v, c)- f.

    -

    , :

    1) (< x1, . . . , xm >) =m

    i=1 xi/m,2) (< x1, . . . , xm >) =

    mi=1 xi,

    3) s(< x1, . . . , xm >) = max{xi i = 1, . . . ,m}.4.

    [3] ( ,

    . 1953., , , [5]). -

    [4], W rV , v W T v W , T - (Q,B), v v (e) = v((e)), T , e B. , rV n, rV (n), rpV - ( , , -

    -

    ). , Supp v

    v : Supp v = {R B v(e R) = v(e), e B}. 10. W rV : W rV 1, :

    A1. (v) o 0, v W rV+;A2. ( v) = (v), T , v W ;A3. (v)(R) = v(Q), R Supp v, v W.

    v(Q) (., , [5]), - - - v(Q), e B. C(v) v

    C(v) = { rV 1 (Q) = v(Q), (e) v(e), e B}.2

    , m - < x1, . . . , xm > xi R.

    24

  • C(v) 1962 ... [6]. , , -

    : v V , e, e B v(e e) + v(e e) v(e) + v(e). , -

    Hv v, -

    Hv(f) = sup {fd

    C(v)}, f I(Q,B), .

    11 [1]. rpV B Sh : rpV B R,

    Sh(v, f) =fdv, v rpV, f B.

    , , -

    ,

    , -

    .

    12 [3]. -

    rpV B P : rpV B R,

    P (v, f) =fdv, v rpV, f B. [3], fV -: fV = {v V | R Supp v : (|R|

  • I(Q,B)n, -

    P v (f, . . . , f) = Pv(f), f I(Q,B).

    3. v rV (n). e B

    (v)(e) = P v (e, Q, . . . , Q).

    ,

    -

    .

    4. v rpV

    Hv(f) =fsdv, f I(Q,B).

    , -

    , ( -

    , , ..).

    05-02-02005a 07-06-00363.

    1. .. . -

    . // . 1998. . 1, 2. . 24-67.

    2. .. . . .: ,

    1961.

    3. .. . a . // -

    . 1975. . 16(33). . 99-120.

    4. . , . . . .: ,1977.

    5. . . . .: , 1974.

    6. .. . n . // , . ., ., .1962. . 13. . 141-142.

    7. . , . . . .: ,1962.

    ,

    . .. , - , 4,

    , 630090, , . (383) 333-26-83, (383) 333-25-98,

    E-mail: [email protected]

    26

  • . .

    1. . ,

    Rk

    .

    NP-, ,

    , -

    .

    -

    .

    , , , -

    , , , -

    . [2,4]

    k- Rk,

    . . x =x21 + . . .+ x

    2k.

    :

    1: V = {~v1, ~v2, . . . , ~vn} - Rk m < n. V m, .

    1 :

    ki=1

    ( nj=1

    vijxj)2 max; (1)

    nj=1

    xj = m; (2)

    xj {0, 1}, 1 j n. (3)

    2: V = {~v1, ~v2, . . . , ~vn} Rk m l, lm < n. V X = {~va1 , ~va2 , . . . , ~vam}, ai+1 ai l i = 1, 2, . . . ,m 1. k l , [2, 4].

    2. 1 2.

    1. [1] 1 2 NP-.

    2. [1] 1 -

    ,

    k18L2

    O(nk2(2L+ 1)k1

    ),

    27

  • L .

    3. [1] k Rk 1 L = L(n), L(n) - n. k Rk - . ,

    = k18L2

    . L = (k18)1/2 -

    O(nk2

    (k 12

    + 1)k1)

    .

    4. [1] 2 -

    ,

    k18L2

    ,

    O(nk(k +m)(2L+ 1)k1

    ).

    b - V.

    5. [1] k Rk , 1 c

    L = 0, 5kmb O(nk2(kmb)k1).

    3.

    1 c .

    3.1. .

    .

    ~B Zk1+ Bi =m

    r=1 vi,r(i), 1 i < k, (i) , (vi1, vi2, . . . , vin) i-

    (vij) ; B = {~ Zk1+ 0 ~ ~B}. 6. fmn(~) { n

    j=1

    vkjxj nj=1

    vijxj = i, 1 i < k;nj=1

    xj = m; xj {0, 1}, 1 j n}.

    1 c -

    S = max{k1i=1

    2i + f2mn(

    ~) ~ B}.. 1 (1)-(3)

    k1i=1

    2i +( nj=1

    vkjxj)2 max; (4)

    nj=1

    vijxj = i, 1 i < k; (5)

    28

  • ~ B; (6)nj=1

    xj = m; (7)

    xj {0, 1}, 1 j n, (8) .

    A {fmn(~) ~ B}. < m,n; ~ > fmn(~)

    {< , j; ~ > 1 j n; 0 ~ ~B} f,j(~).

    f,j(~) < , j; ~ >, - ~v1, . . . , ~vj ~vj.

    A.

    :

    a) f1,j(~) :=; f1,j(~) := j = 1, . . . , n ~ B.b) f1,j(~vj) := vkj j = 1, . . . , n.

    c) f1,j(~) := max{f1,j(~); f1,j1(~)} j = 1, . . . , n ~ B. = 2, . . . ,m :

    a) f,j(~) := j = , . . . , n ~ B.b)

    ~ B

    f,j(~) =

    {f,(~) j = ,

    vk,j + f,j1(~ ~vj), < j n;

    f,j(~) =

    {f,(~) j = ,

    max{f,j(~); f,j1(~)}, < j n.

    A

    7. k Rk , 1 c

    O(mn

    k1i=1 Bi

    ).

    3.2. ~v1, . . . , ~vn - .

    A 1 ~v1, . . . , ~vn .

    Bi =m

    r=1(vi,r(i) vi,r(ni+1)), 1 i < k,

    ~B Zk1+ ; B = {~ Zk1+ | 0 ~ ~B}; bi = min{vij |1 j n} 1 i k.

    29

  • vij = vij bi, (vij) c . A 1 - (vij), (~v1, . . . , ~vn) :

    S = max{k1i=1

    (i mbi)2 +(fmn(~)mbk

    )2 ~ B}., 6

    . ,

    ~B b ( V ):

    8. k Rk , 1 c

    O(mn(mb)k1

    )( ) O

    (mn(2mb)k1

    )(

    ).

    .

    1) -

    2.

    2) 1 2

    Rk.3)

    ( 04-77-7173) .. ,

    .. .. [1].

    4)

    .. [4].

    ( 05-01-00395, 07-07-00022, 07-

    07-00168).

    1. . . , . . , . . , . . . -

    // . .

    , 2, 2007, . 14, N 1.

    2. .. , .. , .. , .. . -

    -

    // . . . .

    2006. T. IX, N(25). . 5574.

    3. . , . . M.:

    . 1982.

    4. .. , .. , .. . -

    - -

    // . . . . 2002. . V, N 2(10). . 94-108.

    , . .. ,

    . , 4, , 630090, , . (8-383-3) 33-21-89,

    (8-383-3) 33-25-98, E-mail:[email protected]

    30

  • . . , . .

    (), , -

    . (-,

    , ) -

    . - -

    , . -

    ,

    .

    . ,

    . -

    , -

    .

    , -

    . ,

    , -

    .

    , , ,

    . -

    -

    . -

    MATLAB.

    (

    ) ( ).

    P-IV 2.6 -

    . .

    (, CPLEX) -

    MATLAB

    -

    . .

    8

    30 .

    06-01-00547

    -2240.2006.1.

    , .. ,

    . 40, , 119991, , . (495) 135-00-20. E-mail: [email protected]

    , .. ,

    . 40, , 119991, , . (495) 135-61-61. E-mail: [email protected]

    31

  • . . , . . , . .

    -

    (),

    -

    [1].

    G = (V,E) V - E. (i, j) E lij, - () cij, rij Qij. S. s = 1, . . . , S V s V . -

    . i V s csi . 0

    s, -

    rs0. [2].

    T 0. Pk 0 k T , Tj T j, Cj =

    (i,j)Tj cij +

    iTj ci Tj. (i, j) T

    dij = rij

    (cij2+ Cj

    ). (1)

    0 k T tk = r0C0 +

    (i,j)Pk dij.

    s T s, V s, (i, j) E Qij , . -

    NP- [3].

    .

    MAD s Qs -. - Qs, s = 1, . . . , S, , -

    .

    .

    .

    ,

    , , -

    . -

    .

    32

  • .

    , ,

    .

    MAD

    G = (V , E ), V . (i, j) E dij (1), G

    , -

    q, MAD.

    MAD.

    MAD

    0. T = (0, ), t0 = 0. 1. (i, j) = arg min

    (u,v)E; uT, v /T{tu(T {(u, v)}) + duv},

    tu(T {(u, v)}) = tu(T ) + r0(cuv + cv) +ePu

    re(cuv + cv).

    T = T {(i, j)} tk, k T . j T , tj = ti+dij,

    tk T . j , k T , - k T , - .

    .

    T , 1.

    MAD ,

    .

    G, q dij. - G , , , G . dij - Cj Tj. , dij , - MAD .

    MAD , -

    dij . q, .

    , . . Qs. s - Qs. i I = E G j J = Ss=1Qs . aij = 1, i j, aij = 0 . xj j xj = 1, j , xj = 0 . :

    iI

    (max

    {0,jJ

    aijxj qi})2 min

    xj{0,1};

    jQsxj = 1, s = 1, . . . , S.(2)

    33

  • qi, i I. ,

    (2). -

    .

    f(y) =

    iI(max{0, yi qi})2, yi =

    jJ aijxj. 0. , , xj = 1/|Qs|, j Qs, s = 1, . . . , S. yi =

    jJ aijxj, i I, L .

    1. gi = max{0, yi qi}, i I f(y) = iI(gi)2. - f(y),

    f(y) f(y) +f(y)(y y) = f(y) +iI

    2gi(yi yi),

    y . ,

    iIgiyi =

    iI

    gijJ

    aijxj =jJ

    (iI

    giaij)xj min

    x.

    s.

    minx

    jJ

    (iI

    giaij)xj =

    Ss=1

    minjQs

    (iI

    giaij).

    GY =

    iI giyi;GAj =

    iI giaij, js = argminjQs GAj GZ =S

    s=1GAjsGY . f(y) f(y) + 2GZ. L = max{L, f(y) + 2GZ}. f(y) L max{1, L}, > 0 , .

    = (j)

    j =

    {1 xj, j = js;xj, j 6= js; j Qs, s = 1, . . . , S;

    z = (zi) zi =

    jJ aijj, i I. , x x = (xj), xj = 1, j {j1, . . . , jS}, xj = 0 . t (0, 1] h(t) =

    f(y+ zt) =

    iI(max{0, yi qi+ zit})2, r(t) =

    iI(gi+ zit)2. , h(0) = r(0), h(0) = r(0) h(t) r(t) t [0, 1]. t = argmin{r(t)|t [0, 1]} = min{1,GZ/ZZ}, GZ = iI gizi ZZ =

    iI(zi)2. y , h(0) < 0 , ,

    r(0) < 0. t > 0. h(0) > h(t) h(0) =r(0) > r(t) h(t), . . t . , xj = xj + j t, j J ; yi = yi + zit, i I, 1.

    34

  • . -

    O(|I| |J |), O(1 ln 1).

    -

    , 1900 3500 .

    2 20. G 4100 7400. .

    -

    05-01-00395.

    1. J. Hu, S. S. Sapatnekar. A Survey on Multi-net Global Routing for Integrated Circuits.

    // Integration, the VLSI Journal. 2002. V. 31. P. 149.

    2. J. Rubinstein, P. Peneld, M. A. Horowitz. Signal Delay in RC Tree Networks. // IEEE

    Trans. on CAD. 1983. V. 2. P. 201211.

    3. M.R. Kramer, J. van Leenwen. Wire-Routing is NP-Complete. // Technical Report

    RUU-CS-84-4, Department of Computer Science, Rijksuniversiteit Utrecht. 1982.

    , . .. ,

    . , 4, 630090, , , . (8-383) 333-37-88, (8-383)

    333-25-98, E-mail: [email protected]

    , . ..

    , . , 4, 630090, , , . (8-383) 333-21-89,

    (8-383) 333-25-98, E-mail: [email protected]

    , . .. ,

    . , 4, 630090, , , . (383) 333-37-88, (8-383) 333-

    25-98, E-mail: [email protected]

    35

  • ..

    ,

    . -

    [1]. ,

    ,

    , -

    [2].

    -

    .

    (

    ) . -

    .

    j = 1, . . . , n n 2. Rn+, R

    n++ -

    .

    . -

    Q(P, v, U) = argmax

    U(Q) : Q Rn+,nj=1

    PjQj = v

    , (1) P Rn++, v 0 - U . U Rn+ , - Q(P, v, U) (1) v., Rn+ , [3].

    U , P Rn++, v 0, 0

    Q(P, v, U) = Q(P, v, U).

    U, U , P Rn++, v 0

    Q(P, v, U) = Q(P, v, U).

    U i i = 0, . . . , k k 2. i = 1, . . . , k , i = 0 . , P Rn++, vi 0, i = 1, . . . , k

    Q(P,ki=1

    vi, U i) =ki=1

    (Q(P, vi, U i).

    , . ..

    , , 130, , 664033, , . (8-3952) 42-88-27,

    (8-3952) 42-67-96, E-mail:[email protected]

    36

  • , -

    , . [4, 5]

    [6]

    1. -

    , -

    .

    C -

    -

    . ,

    -

    . ,

    . ,

    - .

    . -

    . Itp , Itq

    j = 1, . . . , n, t , t > . - ,

    [7-10]. ,

    Rn++:

    Itp = f(P , Q , P t, Qt), Itq = (Q

    , P , Qt, P t).

    P , Q , P t, Qt , - , t, . Rn++ R

    1+.

    -

    , -

    . f , [8], .

    1. ( ) -

    Itp Itq =nj=1

    P tjQtj/

    nj=1

    P j Qj . (2)

    2. () -

    ()

    maxj=1,...,nPtj /P

    j Itp minj=1,...,nP tj /P j , (3)

    maxj=1,...,nQtj/Q

    j Itq minj=1,...,nQtj/Qj . (4)3. , t, l

    Itp I tlp = Ilp , (5)

    37

  • Itq I tlq = Ilq . (6) [11, 12]

    2. (2) (6) . f , . -

    ( ) -

    . -

    . [13] , .

    [14].

    P (s), Q(s) - s Rn++, - j = 1, . . . , n. - , t

    Dtp = exp t

    nj=1

    Qj(s)

    V (s)dPj(s), D

    tq = exp

    t

    nj=1

    Pj(s)

    V (s)dQj(s),

    V (s) =nj=1

    Pj(s)Qj(s)

    t.

    Dtp Dtq = V (t)/V ()

    . -

    . , -

    [15],

    P (t) = P (), Q(t) = Q()

    Dtp > 1, Dtq < 1.

    , ,

    [, t] 1. "" - , ,

    . [, t] , .

    [10, 15, 16] , -

    , , ,

    (1). , [15], ,

    , -

    .

    38

  • U , P (s) - Rn++, v(s) s [, t]. s [, t] -

    Q(s) = argmin

    U(Q) :nj=1

    Pj(s)Qj = v(s), Q Rn+ . (7) P (s), Q(s) s [, t] - U , Q(s) (7) v(s) > 0 s [, t]. U - , ,

    , P (s), Q(s) P (s), Q(s), s [, t] , ..

    P () = P (), Q() = Q(), P (t) = P (t), Q(t) = Q(t),

    Dtp = Dtp , D

    tq = D

    tq .

    : Dtp , Dtq P (s), Q(s); D

    tp , D

    tq

    P (s), Q(s).C

    3. U ,

    .

    06-02-00266 .

    1. . . . , . .: , 1948.

    2. . , . . . .: , 1971.

    3. . . . .: , 1983.

    4. .. . : -

    ? .: ,

    1997.

    5. .. .

    // -

    . . .: , 1994.

    6. .. . . .:

    , 2000.

    7. .. . . .: , 1963.

    8. . . . .: , 1928.

    9... . ( -

    ). .: , 1992.

    10. . . . .: , 1980.

    11. .. . . .: ,

    39

  • 1991.

    12. .. . // -

    , 1993. 2.

    13. Devisia. Economic rationelec. Paris, 1928.

    14. .. . //

    , 1929. 9-10.

    15. .. . : ,

    1996.

    16. .. . . .: ,

    2004.

    40

  • . .

    1.

    U A 2U , : A A, A A A A. S = (U,A) U . A , - U . D. , D : D D, D D D D. A, D S, S = (U,A) S = (U,D) , .

    S -, .

    :

    max{f(X) : X B}, (1)min{f(X) : X C}, (2) B , C S = (U,A) = (U,D), f : 2U R+ .

    -

    ,

    , ,

    p-, , k- - . ,

    NP-.

    , (1)

    (2), , -

    . , f : 2U R+ , - X, Y U f(XY )+f(XY ) f(X)+f(Y ), , . -

    f . , f() = 0 , . (1)

    GA ( ).

    0. X0 , 1. i (i 1). xi / Xi1, f(Xi1 {xi}) = max

    x/Xi1,Xi1{x}A

    f(Xi1 {x}).

    Xi Xi1 {xi}, i+ 1. xi / Xi1 , GA Xi1..

    (2)

    .

    41

  • GR.

    0. X0 U , 1. i (i 1). xi Xi1, f(Xi1 \ {xi}) = min

    xXi1,Xi1\{x}D

    f(Xi1 \ {x}).

    Xi Xi1 \ {xi}, i+ 1. xi Xi1 , GR Xi1..

    , GA , GR -

    , . . GA (1), GR (2).

    2.

    .

    S = (U,A) = (U,D) W U . W , W . W , W .

    cA(S) = minWU,W /A

    rmin(W )

    rmax(W ), cD(S) = max

    WU,W /D

    gmax(W ) |W |gmin(W ) |W | ,

    rmin(W ) rmax(W ) W , gmin(W ) gmax(W ) W , . , cA(S) 1 cD(S) n 1 S. S , cA(S) = 1, -, cD(S) = 1. rmax(U) , gmax() - . p- S = (U,A) p- S = (U,D), A = {A U : |A| p}, D = {D U :|D| p}, p , p < |U |. .

    1 (-) [5, 12]. S = (U,A) - U . GA (1) S f : U R+ , S . -

    . (2) , -.

    2. S = (U,D) . GR - (2) S , S . [6].

    1, ,

    (1)

    . [8, 9] GA

    :

    f(GA)

    f(OPT ) cA(S), (3)

    42

  • OPT (1). [6] GR -

    (2) :

    f(GR)

    f(OPT ) cD(S). (4)

    -

    , (3) (4), GA GR

    (1) (2) .

    , [2] , (2) -

    ,

    -

    , (2).

    3.

    (1) , -

    .

    ,

    . -

    (1) ,

    , -.

    U , f : 2U R+. V U |V | 3 k |V | 2. X = {x1, . . . , xk} V GA- V ,

    f({x1}) f({x}) x V ,f({x1, x2}) f({x1, x}) x V \ x1,. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .f(X) f({x1, . . . , xk1, x}) x V \ {x1, . . . , xk1}. f : 2U R+ GA-, V U |V | 3 GA- X ={x1, . . . , xk} f(V \ xk) f(V \ x) x V \X., GA-. - (1).

    3. S = (U,A) . GA - (1) S GA-- f : 2U R+ , S . -

    (1) .

    4. f : 2U R+ . - GA (1) f S = (U,A) , f - GA-. (2): -

    (2) , GR ( GR- ), 2 (2)

    .

    43

  • 4.

    (1), B p, f : 2U R+ , f() = 0. (1) p- f(X) =

    jJ

    maxiX

    cij, (cij) n m U J . [4] GA p- :

    f(GA)

    f(OPT ) 1

    (p 1p

    )p e 1

    e 0, 63. (5)

    [10] (5) -

    p- . [3] :

    f(GA)

    f(OPT ) 1

    c

    (1

    (p cp

    )p),

    c [0, 1] f , - . c :

    c = maxxU,

    f({x})>f()

    f({x}) f() (f(U) f(U \ {x}))f({x}) f() ,

    c = 0 , f ( f() = 0). (2), C - p, f : 2U R+ ,f(U) = 0. p- - f(X) =

    jJ

    miniX

    cij, (cij) nm U J . , f() = max

    X,YU,XY=

    {f(X) + f(Y ) f(X Y )}

    p- . , GR

    . , ,

    , p- , , ,

    P = NP [11]., (2)

    GR.

    f

    s = maxxU,

    f({x})

  • [1] s < 1 - GR

    p- :

    f(GR)

    f(OPT ) 1

    t

    ((q + t

    q

    )q 1

    ) e

    t 1t

    ,

    q = n p, t = s/(1 s). [7] (2) -

    p. GR

    p- (cij).

    1. .. .

    // . . . C. 1. 1998. . 5,

    N 4. C. 45-60.

    2. .. , .. . // .

    . . C. 1. 2003. . 10, N 3. C. 54-66.

    3. M. Conforti, G. Cornuejols. Submodular set functions, matroids and the greedy al-

    gorithm: Tight worst-case bounds and some generalizations of the Rado-Edmonds the-

    orem // Discrete Appl. Math. 1984. V. 7, N 3. P. 251-274.

    4. G. Cornuejols, M.L. Fisher, G.L. Nemhauser. Location of bank accounts to optimize

    oat: An analytic study of exact and approximate algorithms // Management Science.

    1977. V. 23. P. 789-810.

    5. J. Edmonds. Matroids and the greedy algorithm // Math. Programming. 1971, V. 1,

    N 2. P. 127-136.

    6. V. Il'ev. Hereditary systems and greedy-type algorithms // Discrete Appl. Math. 2003.

    V. 132, N 1-3. P. 137-148.

    7. V. Il'ev, N. Linker. Performance guarantees of a greedy algorithm for minimizing a

    supermodular set function // European J. Oper. Res. 2006. V. 171, N 2. P. 648-660.

    8. Th.A. Jenkyns. The ecacy of the greedy algorithm // Proc. 7th S-E Conf. Combi-

    natorics, Graph Theory and Computing. 1976. P 341-350.

    9. B. Korte, D. Hausmann. An analysis of the greedy heuristic for independence systems //

    Annals of Discrete Mathematics. 1978. V. 2. P. 65-74.

    10. G.L. Nemhauser, L.A. Wolsey, M.L. Fisher. An analysis of approximations for maxi-

    mizing submodular set functions I // Math. Programming. 1978. V. 14. P. 265-294.

    11. G.L. Nemhauser, L.A. Wolsey. Integer and combinatorial optimization. New York.:

    John Wiley & Sons, Inc., 1988.12. R. Rado. Note on independence functions // Proc. London. Math. Soc. 1957. V. 7,

    N 3. P. 300-320.

    ,

    , . , 55, , 644077, , .

    (3812) 22-56-96. E-mail: [email protected]

    45

  • NP-

    . .

    , -

    , -

    -

    .

    ,

    ( ) (o-line) -

    . -

    , ()

    , -

    (); . [1-3] .

    ,

    q, - :

    xn =mM

    unnm(m), n = 0, . . . , N 1,

    unnm(m) = 0, n nm 6= 0, . . . , q 1; (u0(m), . . . , uq1(m)) Rq,0 < (u0(m), . . . , uq1(m))

  • X X(n1, . . . , nM , U1, . . . , UM), -

    .

    () -

    ( X() Y X(),2I), . -

    Y X()2 .

    Yn = (yn, . . . , yn+q1), n = 0, . . . , N q + 1.1. . -

    , Um = U = (u0, . . . , uq1), m M, .. ,

    :

    xn =mM

    unnm , n = 0, . . . , N 1.

    X = X(n1, . . . , nM , U) .1.1. ; .

    : Y RN , U Rq M . : (n1, . . . , nM) M ,

    mM(Ynm , U) max .

    . , -

    O[M(Tmax Tmin + q)(N q + 1)] = O(MN2), [4].1.2. ; .

    : Y RN U Rq. : (n1, . . . , nM) , mM

    {2(Ynm , U) U2} max .

    , O[(TmaxTmin + q)(N q + 1)] = O(N2), [5].1.3. ; .

    : Y RN , M q. : (n1, . . . , nM) M,

    mM

    Ynm max .

    NP-. ,

    O[M(Tmax Tmin + q)(N q + 1)] = O(MN2), [6].1.4. ; .

    : Y RN , q. : (n1, . . . , nM) ,

    1

    MmM

    Ynm2 max .

    47

  • . , NP-.

    2. -

    -. -

    :

    xn =mL

    unnm +

    mM\Lunnm(m), n = 0, . . . , N 1,

    L M, |L| = L. U , Um,m M\L, . X = X(n1, . . . , nM , U,L, {Um,m M\L}) .

    2.1. .

    2.1.1. . -

    , Um {U : U Rq, 0 < U < },m M \ L. .2.1.1.1. .

    : Y RN , U Rq, M L. : (n1, . . . , nM) M L ,

    mL2(Ynm , U) +

    mM\L

    Ynm2 max .

    2.1.1.2. , .

    : Y RN , U Rq, L. : (n1, . . . , nM) L - , EYnm 6= 0 ( E ), m M \ L.2.1.1.3. , .

    : Y RN , U Rq, M . : (n1, . . . , nM) M L ,

    mL{2(Ynm , U) U2}+

    mM\L

    Ynm2 max .

    2.1.1.4. .

    : Y RN U Rq. : (n1, . . . , nM) L, , , EYnm 6= 0, m M \ L.2.1.2. . -

    , U A Um A, m M \ L, A {U : U Rq, 0 < U < }, |A| = K, .. - () A . - .

    2.1.2.1. .

    : Y RN , A K, M L. : (n1, . . . , nM) M , L {Um A \ {U},m M \ L},

    mL2(Ynm , U) +

    mM\L

    {2(Ynm , Um) Um2} max .

    48

  • 2.1.2.2. , .

    : Y RN , A K, L. : (n1, . . . , nM) , L {Um A \ {U},m M \ L}, .

    2.1.2.3. , .

    : Y RN , A K, M . : (n1, . . . , nM) M , L {Um A \ {U},m M \ L},

    mL{2(Ynm , U) U2}+

    mM\L

    {2(Ynm , Um) Um2} max .

    2.1.2.4. .

    : Y RN , A K. : (n1, . . . , nM) , L {Um A \ {U},m M \ L} , .

    2.2. .

    2.2.1. . -

    1

    LmL

    Ynm2 +

    mM\L{2(Ynm , Um) Um2} max,

    . NP-

    ( NP- 1.3).

    2.2.1.1. .

    : Y RN , A K, M L. : (n1, . . . , nM) M , L {Um A,m M \ L}.2.2.1.2. , .

    : Y RN , A K, L. : (n1, . . . , nM) , L {Um A,m M \ L}. . ,

    NP- ( 1.4).

    2.2.1.3. , .

    : Y RN , A K, M . : (n1, . . . , nM) M , L {Um A,m M \ L}.2.2.1.4. .

    : Y RN , A K. : (n1, . . . , nM) , L {Um A,m M \ L}.2.2.2. . -

    1

    LmL

    Ynm2 +

    mM\LYnm2 max,

    .

    2.2.2.1. .

    : Y RN , M L. : (n1, . . . , nM) M L.

    49

  • 2.2.2.2. , .

    : Y RN , L. : (n1, . . . , nM) L , EYnm 6= 0, m M \ L. NP- ( NP- -

    1.3). .

    , NP- ( 1.4).

    2.2.2.3. , .

    : Y RN , M . : (n1, . . . , nM) M L.2.2.2.4. .

    : Y RN . : (n1, . . . , nM) L , EYnm 6= 0, m M \ L.

    -

    . -

    , NP-, -

    . -

    ,

    .

    06-01-00058 07-07-00022.

    1. Kel'manov A.V., Jeon B. A Posteriori Joint Detection and Discrimination of Pulses in

    a Quasiperiodic Pulse Train // IEEE Trans. on Signal Proc. 2004. Vol. 52, No. 3 P. 1-12.

    2. .. -

    :

    // XII -

    (-12). , 2005. . 125-128.

    3. .. -

    -

    // 3-

    . , 2006. . 37-41.

    4. .., ..

    //

    . 2001. .41, 5 . 807-820.

    5. Kel'manov A.V., Khamidullin S.A., Okol'nishnikova L.V. A Posteriori Detection of

    Identical Subsequences in a Quasiperiodic Sequence // Pattern Recognition and Image

    Analysis. 2002. Vol. 12, No. 4. P. 438-447.

    6. .., .., .., .. -

    // . . . . 2006. .9, 1(25).

    . 55-75.

    ,

    . .. ,

    . 4, , 630090, ,

    .: (8383) 333-3291, : (8383) 333-2598, e-mail: [email protected]

    50

  • . . , . .

    -

    , , -

    , [2, 4, 22, 23].-

    NP- . -

    .

    , ,

    . -

    , ,

    . -

    (), p- ( ), ,

    , [2, 4, 6, 1113, 16, 18, 24].

    -

    (). ,

    p- , : . ,

    ,

    [2, 10, 11, 13, 14, 20].

    -

    . ,

    "" , ,

    .

    1.

    . p- ( Pmin) . - I = {1, . . . ,m} , , J = {1, . . . , n}. . -

    i- j- cij, i I, j J . p, ,

    , -

    .

    : zi = 1, i- - , zi = 0, i I;xij = 1, j- i- , xij = 0 , i I, j J .

    51

  • :

    f(z, x) =iI

    jJ

    cijxij min (1)

    iIzi = p, (2)

    iIxij = 1, j J, (3)

    xij zi, i I, j J, (4)xij, zi {0, 1}, i I, j J. (5) z = (z1, . . . , zm) (1)-(5), (2) (5).

    cij dij i- j- . p- (Pmax): p , .

    Pmin , , .. (2),

    c0i , i I, :

    iIc0i zi +

    iI

    jJ

    cijxij min .

    -

    . , -

    ,

    [11].

    . -

    -

    , , c0i i- , cij j- , i- , i I, j J. j i, dij .

    .

    [6].

    , -

    , [3].

    [15].

    , NP-.

    2.

    Pmin. -

    .

    52

  • , F (k) k- , F (0) =.

    D (1) = . k (k 1) 1. z(k) (k). - , : ,

    F (k1), . 2. T (z(k)) - z(k):

    iIjJ

    cijxij min

    iIxij = 1, j J,

    0 xij z(k)i , i I, j J.,

    , x(k), - xij . f(z

    (k))

    T (z(k)). F (k) = min{f(z(k)), F (k1)} - .

    3. ( ):

    (k)1 z1 + . . .+

    (k)m zm (k)0 . (6)

    (k+1) (k), - (6). .

    (6) , :

    a) z(k);b) z (k), f(z) < F (k). "a" D, "b" .

    "a", , ,

    .

    , "a" "b", -

    iIk0

    zi 1, (7)

    Ik0 = {i I : z(k)i = 0}. - [9]. , (7)

    (k) z(k). , D Cpm.

    53

  • (6) [17], -

    . 1 z(k) - T (z(k)).

    jJuj

    iIjJ

    wijz(k)i max

    uj wij cij, wij 0, i I, j J.(8)

    , (-

    ) u(k)j , w

    (k)ij , i I, j J . , z(k) , :

    iI

    jJ

    w(k)ij zi >

    jJ

    u(k)j F (k). (9)

    (8) ,

    z(k) . , .. -

    , (k). [14] p- - .

    .

    3.

    Pmin - , , , 1, M (M 2). C. , (9) -

    1.

    , D - . -

    (7).

    , -

    , z , - [10].

    p-, D . Pmin (m m)- , - 0, (,

    m p+ 1m p ) > 0. , Pmin Cpm D -. Pmin , NP-. Pmax [10,24].

    ,

    C,

    54

  • c0i = 1, i I. z(k), , 1 [10].

    [8, 21, 25]. , [7, 21]

    L- - . , (

    ) . [25] , D (

    ), -

    ,

    z(1), . . . , z(K) . [11, 13, 14] , -

    , p- . -

    L- , , - .

    OR-Library [19], TSPLIB [26], [5],

    .

    -

    .

    1. .. : -

    // . "

    ", , 1998. . 410.

    2. .., .., .. -

    . : , 1978. 160 .

    3. .. -

    . : - , 2005. 408 .

    4. .., .., .. -

    . , 1978. 333 .

    5. .., .., .., .. -

    // XII . -, , 2001. . 1. . 132137. http://math.nsc.ru/AP/benchmarks/

    6. .., .., .. -

    . . : ,

    1997. 26 .

    7. .., .. -

    // . 2004. 3.

    . 4854.

    8. ., .., .. -

    // . 2004. 2.

    . 7992.

    9. .. -

    // . 1994. 2. . 1839.

    10. .., .., ..

    55

  • //

    . 2005. 2(31). . 7680.

    11. .., ..

    : -. . : , 2004. 40 .

    12. .., .., ..

    // . 2006.

    4(38). . 62-67.

    13. .., ..

    // XII -

    " ". , , 2001. . 1.

    . 207-210.

    14. ..

    p-. . : , 2006. 23 .15. .., .., .. , -

    // III "

    ". , 2006. . 105.

    16. .., .. -

    p- // . 2004. 3. . 8088.17. . : 2- . .:

    , 1991. 702 .

    18. Avella P., Sassano A., Vasil'ev I. Computational study of large scale p-median problems//Mathematical Programming. 2006. Vol. 109. 1. P. 89114.

    19. Beasley J.E. A note on solving large p-median problems //European Journal of Opera-tional Research. 1985. 21. P. 270273.

    20. Benders J. F. Partitioning procedures for solving mixed-variables programming problems

    //Numerische Mathematik. 1962. Vol. 4. .3. P. 238252.

    21. Devyaterikova M.V., Kolokolov A.A. On the stability of some integer programming

    algorithms //Operations Research Letters. 2006. 34(2). P. 149154.

    22. Discrete Location Theory (Ed. by Pitu B. Mirchandani and Richard L. Franscis), by

    John Wiley and Sons, Inc., 1990. 576 p.

    23. Facility Location. Applications and Theory (Ed. by Zvi Drezner and Horst W. Hamacher),

    Springer, 1990. 457 p.

    24. Kolokolov A.A., Kosarev N.A. Analysis of decomposition algorithms with Benders

    cuts for p-median problem //Journal of Mathematical Modelling and Algorithms. 2006. 5. P. 189199.

    25. Kosarev N.A., Kolokolov A.A. On stability of some benders decomposition algorithms

    for p-median problem //Proceedings of European Chapter on Combinatorial OptimizationXVIII. Minsk, 2005. P. 26.

    26. Reinelt G. TSPLIB: A traveling salesman problem library //ORSA Journal on Compu-

    ting. 1991. 3. P. 376384.

    http://www.iwr.uni-heidelberg.de/groups/comopt/software/TSPLIB95/.

    , ,

    . .. ,

    . , 13, , 644099, , . (3812) 23-67-39, (3812) 23-45-84.

    E-mail: [email protected], [email protected]

    56

  • . .

    -

    . ,

    , . -

    . -

    T = {T1, . . . , Tn}, h(Tj) w(Tj) Tj, C = {C1, . . . , Cm} , wi i- . -

    T ( , - ), ,

    -

    .

    , NP-,

    m [1]. .

    HO(T,C) HA(T,C) , A. , 1.

    A

    RA = supT,C{HA(T,C)/HO(T,C)},

    RA = limk

    supT,C{HA(T,C)/HO(T,C) | HO(T,C) k}.

    . -

    , T , - , -

    T C - - .

    -

    . [2].

    [3] -

    , 1.7. [4]

    ,

    , 1.69103, ,

    . [5] ,

    -

    , 1.5401.

    [21] A RA 1.5889.

    57

  • -

    m , , m 1

    1. [6] , -

    .

    2 1/m, m . [8] - 1.986

    m > 70. 1.945, 1.923 [10] [7]. [11] ,

    , 1.837

    m. [7] 1.852.

    . [12] , -

    (

    ),

    10.

    e . - ...

    1. A : RA e. r , 0 < r < 1.2. Ar , R

    Ar 2er .,

    [13-21].

    -

    . [4] -

    . r (0, 1). , . -

    R, k, rk+1 < h(R) rk rk. - ,

    rk. - . , ,

    ( k), - , .

    A(E) ,

    E.

    U([0, 1]) [0, 1]. - , hi wi [0, 1]. , wi, hi . - , ,

    . ,

    N (N ). E

    E

    (L

    Ni=1

    wi

    )= O (f(N)) ,

    58

  • L - , E {wi}.. [20] E f(N) = N log N, 0 < < 1, 0. A(E) :

    = O

    (N

    (log N

    1+ N1

    )).

    = N (1)/(2) log(/(2))N ,

    = O(N1/(2) log(/(2))N

    ).

    , , -

    , FF (rst t): f(N) = N2/3, BF (best t): f(N) = N1/2(logN)3/4, (best on-line) BO: f(N) = (N logN)1/2 [16]. , A(FF) =O(N3/4), A(BF) = O(N2/3(logN)1/2) [19], A(BO) = O(N2/3(logN)1/3). [17].

    , ,

    on-line, , [14], -

    = O(N1/2 logN

    ).

    on-line . , -

    ( )

    . -

    : = (N2/3

    ). 05-01-00798.

    1. Garey M.R., Johnson D.S., Computers and Intractability, A Guide to the Theory of

    NP-Completeness, Freeman, San Francisco, 1979.

    2. Baker B.S., Coman E.J., Rivest R.L. Orthogonal packings in two dimensions// SIAM

    J. Computing, 1980, V. 9, P. 846-855.

    3. Baker B.S., Schwarz J.S. Shelf algorithms for two dimensional packing problems//

    SIAM J. Computing, 1983, V. 12, P. 508-525.

    4. Csirik J., Woeginger G.J., Shelf Algorithms for On-Line Strip Packing// Inf. Process.

    Lett., 1997, V. 63, N 4, P. 171-175.

    5. Van Vliet A. An improved lower bound for on-line bin packing algorithms// Inf. Process.

    Lett., 1992, V. 43, P. 277284.

    6. Graham R. L. Bounds for certain multiprocessing anomalies // Bell System Technical

    Journal, 1966, V. 45, P. 15631581.

    7. Albers S. Better bounds for online scheduling// Proc. of the 29th ACM Symp. on

    Theory of Computing, 1997, P. 130-139.

    8. Bartal Y., Fiat A., Karlo H., Vohra R. New algorithms for an ancient scheduling

    problem // J. Comput. Syst. Sci., 1995, V. 51, N 3, P. 359-366.

    9. Faigle U., Kern W., Turan G. On the performance of on-line algorithms for partition

    problems //Acta Cybernetica, 1989, V. 9, P. 107-119.

    10. Karger D. R., Phillips S. J., Torng E. A better algorithm for an ancient scheduling

    problem // Proc. of the 5th Annu. ACM-SIAM Symp. on Discrete Algorithms, 1994, P.

    59

  • 132140.

    11. Bartal Y., Karlo H., Rabani Y. A better lower bound for on-line scheduling // Inf.

    Proces. Letters, 1994, V. 50, P. 113116.

    12. .. -

    // , 2006, . 18, N 1 C. 91105.

    13. E.G. Coman, Jr., C. Courcoubetis, M.R. Garey, D.S. Johnson, P.W. Shor, R.R.

    Weber, M. Yannakakis, Perfect packing theorems and the average-case bahavior of optimal

    and online bin packing // SIAM Review, 2002, V. 44, N 1, P. 95108.

    14. R.M. Karp, M. Luby, A. Marchetti-Spaccamela, A probabilistic analysis of multidimensional

    bin packing problems // Proc. Annu. ACM Symp. on Theorty of Computing, 1984, P.

    289298.

    15. P. W. Shor, The average-case analysis of some on-line algorithms for bin packing//

    Combinatorica, 1986, V. 6, P. 179-200.

    16. P. W. Shor, How to pack better than Best Fit: Tight bounds for average-case on-line

    bin packing // Proc. 32nd Annual Symposium on Foundations of Computer Science, 1991,

    P. 752-759.

    17. E. G. Coman, Jr., D. S. Johnson, P. W. Shor and G. S. Lueker. Probabilistic analysis

    of packing and related partitioning problems // Statistical Science, 1993, V. 8, P. 40-47.

    18. E. G. Coman, Jr., P. W. Shor. Packings in two dimensions: Asymptotic average-case

    analysis of algorithms// Algorithmica, 1993, V. 9, P. 253-277.

    19. .., ..,

    // , 2006, . 18, N 1, C.

    7690.

    20. .., .., -

    // , 2006, . 12, C.

    1726.

    21. Han X., Iwama K., Ye D., Zhang G. Strip packing vs. bin packing// Electronic

    Colloquium on Cpmputational Complexity, 2006, Report N 112.

    ,

    , . , 25, , 109004, , . (8-495) 912-56-59,

    (8-495) 912-15-24, e-mail: [email protected]

    60

  • ..

    . -

    NP-

    [4]. -

    (

    theoretical computer science), ,

    ,

    ,

    .

    ,

    , . , ,

    , ,

    . -

    . " P 6= NP, ..." - [1-3, 6]. , -

    ,

    [9]

    [14],

    [2]. -

    . , -

    [4], .

    , , .

    , -

    P 6= NP . -

    , .

    ,

    .

    . , -

    .

    "" [7], -

    NP- .

    , -

    . , -

    . -

    , ..

    (. [2, 3, 17]). ,

    -

    61

  • "No free lunch"[21].

    ,

    , .

    , -

    , .

    , , -

    , , . -

    -

    [2, 3, 17].

    [2].

    1 ,

    . , -

    .

    1.

    P 6= NP - [1, 17].

    , () -

    , . P 6= NP NP, -

    [17].

    [15]. , ,

    . , -

    , .

    [18] . , -

    ,

    P = NP. , .. . ,

    P 6= NP . , -

    . , P = NP ,

    . , -

    , NP.

    P = NP. - , -

    P = NP, , ,

    NP- . -

    [2], ,

    -

    . ,

    , , P 6= NP. -

    62

  • , , P = NP. [10] . , , -

    ?

    , ..

    .

    , -

    . ,

    .

    [10] ,

    -

    .

    .

    , , .

    . , -

    , .

    ,

    . , , -

    ,

    P 6= NP. , NP - . , -

    .

    -

    , , -

    , .

    . -

    1/2. pM(x) , M x. x . A [2]:1. , pM(x) > 1/2, x A, pM(x) = 0, x 6 A.2. , pM(x) > 1/2+ , x A, pM(x) < 1/2 0, x 6 A, > 0. RP , -

    . BPP

    ,

    . -

    P. , P RP BPP . , RP =NP BPP = NP, P = NP, [2]. , RP 6= NP BPP 6= NP. , NP - .

    1. P 6= NP, - .

    , P 6= NP,

    63

  • .

    2.

    NP-, ,

    . -

    , NP-, -

    . -

    .

    k- [19], -

    [19].

    [5].

    .

    [12], P NP .

    L, . (L, ) [6,8,12]. L. [6,8,12,13,19,20]. B = {0, 1}. B , .

    1. , :B [0, 1], (x) 0 x B

    xB(x) = 1.

    2. f : B N , k, c > 0 ,

    xB

    f 1/k(x)

    |x| (x) c.

    3. L , ( ). (L, ) .

    4. (L, ) AvgP , L . AvgP P - .

    5. P-, - M , x B k 1

    M(x, 1k) (x) 2k.

    6. (L, ) DistNP, L NP P-.

    64

  • DistNP

    NP. P 6= NP DistNP 6 AvgP. .

    7. (L, ) (L, ), - f , x L f(x) L q , y B,

    (y) 1q(|y|)

    xf1(y)

    (x).

    , , (L, ) AvgP, (L, ) . f , , x B

    (x) (x)q(|x|) .

    , L L [1].

    8. (L, ) DistNP-, DistNP

    .

    [6,8,12,20], DistNP- .

    6, [6]. -

    DistNP-

    (K,K) [8,12,20]:: (i, x, 1n), x B, i, n N: Mi x n ? K : i, x n

    K(i, x, 1n) =

    2|i|

    |i|2 2|x|

    |x|2 1

    n2.

    [20]. [13]

    , . -

    NP- ,

    DistNP-.

    NP- [17],

    P- NP-

    , DistNP-.

    9 [13]. L -, q S : 1B 7 B , 1. S .2. x n S(1n, x) L x L.3. x n , n > |x|

    |S(1n, x)| = q(n).

    65

  • q S. S - , .

    10 [13]. L -, E : B B 7 B - D : N B 7 B , 1. S D .2. x, p B E(p, x) L x L.3. |x1| = |x2|, |p1| = |p2| p1 p2,

    E(p1, x1) E(p2, x2).4. |x1| = |x2| |p1| = |p2|, |E(p1, x1)| = |E(p2, x2)| |x1| < |x2|

    |p1| < |p2|, |E(p1, x1)| < |E(p2, x2)|.5. x, p B D(|p|, E(p, x)) = p D(k, w) , x p , |p| = k E(p, x) = w.

    D , .

    2 [13]. L NP- , - - -. P- -

    , L DistNP- .

    DistNP 6 AvgP? , , ,

    (K,K) - K .

    3. DistNP 6 AvgP, - .

    3.

    , DTime(2O(n)) 6= NTime(2O(n)) P6= NP, RP 6= NP DistNP 6 AvgP [2,6]. 1 3 - .

    4. DTime(2O(n)) 6= NTime(2O(n)), -

    .

    , -

    , -

    P-

    . .

    . -

    .

    ,

    [4]. -

    66

  • , -

    , P-

    . -

    P- .

    -

    . ,

    "" NP

    [11]. 2,

    , -

    [13].

    , -

    , -

    . , -

    .

    . ..

    ,

    .

    , -

    , ,

    .

    , , ,

    .

    L [16]. -

    a2t, a t , a < 2L. [16], NP- -

    , L ,

    L . - , .

    06-01-00075.

    1. . , . . . .:

    , 1982.

    2. M. Atallah. Algorithms and theory of computation handbook. Boca Raton; FL: CRC

    Press, 1999.

    3. G. Ausiello, P. Crescenzi, G. Gambosi, V. Kann, A. Marchetti-Spaccamela, M. Protasi.

    Complexity and approximation: combinatorial optimization problems and their approxi-

    mability properties. Berlin: Springer, 1999.

    4. C. Blum, A. Roli. Metaheuristics in combinatorial optimization: overview and conceptual

    comparison.// ACM Computing Surveys. 2003. Vol. 35, No. 3. P. 268-308.

    5. B. Bollobas. Random Graphs. Academic Press, 1985.

    6. S. Ben-David, B. Chor, O. Goldreich, M. Luby. On the theory of average case complexity.//

    J. Comput. System Sci. 1992. V. 44. P. 193-219.

    7. E. Eberbach. Toward a theory of evolutionary computation.// BioSystems. 2005. V.

    67

  • 82. P. 1-19.

    8. Y. Gurevich. Average case completeness.// J. Comput. System Sci. 1991. V. 42. P.346-

    398.

    9. W. E. Hart, R. K. Belew. Optimizing an arbitrary function is hard for the genetic

    algorithm.//Proceedings of the Fourth International Conference on Genetic Algorithms.

    1991. P. 190-195

    10. R. Impagliazzo. A personal view of average-case complexity.// In Proceedings of the

    10th Conference on Structure in Complexity Theory. 1995. P. 134-147.

    11. R. Impagliazzo, L. A. Levin. No better ways to generate hard NP instances than

    picking uniformly at random.// In Proc. of the 31st IEEE Symp. on Foundation of

    Computer Science. 1990. P. 812-821.

    12. L. Levin. Average case complete problems.// SIAM Journal on Computing. 1986. V.

    15, N. 1. P. 285-286.

    13. N. Livne. All natural NPC problems have average-case complete versions.// Electronic

    Colloquium on Computational Complexity, Revision 1 of Report No. 122 (2006). P. 1-17.

    14. P. Moscato. Memetic algorithms: a short introduction// D. Corne, F. Glover, M.

    Dorigo (eds.) New Ideas in Optimization. McGraw-Hill. 1999.

    15. E. Nagel, J. R. Newman, D. R. Hofstadter. Godel's proof. New York University Press,

    2002.

    16. J. Orlin, A. Schulz, S. Sengupta. -optimization schemes and L-bit precision: alternativeperspectives in combinatorial optimization.// Proceedings of the Thirty-Second Annual

    ACM Symposium on Theory of Computing. 2000. P. 565-572.

    17. C. H. Papadimitriou. Computational complexity. Addison Wesley. 1994.

    18. A. A. Razborov, S. Rudich. Natural proofs.// J. Comput. System Sci. 1997. V. 55, N.

    1. P. 24-35.

    19. J. Wang. Average-case computational complexity theory. //L. Hemaspaandra, A.

    Selman (eds.) Complexity Theory Retrospective II. Springer Verlag. 1997. P. 295-328.

    20. J. Wang. Average case intractable NP-problems.// D.-Z. Du, K. Ko (eds.) Advances

    in Complexity and Algorithms. Kluwer Academic Publishers. 1997. P. 313-378.

    21. D. Whitley, J. P. Watson. Complexity theory and the no free lunch theorem.//

    E. K. Burke, G. Kendall (eds.) Search Methodologies: Introductory Tutorials in Optimization

    and Decision Support Techniques. New York: Springer. 2005.

    , . ..

    , . 4, , 630090, , . (383) 333-20-86,

    (383) 333-25-98. E-mail:[email protected]

    68

  • . . ,

    , - . ( s, t ), NP. , - .

    1. -

    . , - [1].

    1.1.

    AS = (X, V, R; P, F, W), - :

    X = (x1, x2,..., xn) ; V = (v1, v2, ..., vg) ; R = (r1, r2, ..., rm) ; P : V 2X , vV

    P(v) X . PS = (X,V; P); F : R VPS2 , rR F(r)

    V . F FS = (V, R; F). - VPS2 , PS, . , .

    rR W : r 2P(F(r)) , rR - W(r) P(F(r)) , P(F(r)) PS, - F(r) V. , W WS = (X, R; W).

    PS AS, WS - AS.

    , AS = (X, V, R), , . :

    P1(x) = {v: xP(v)}; F1(v) = {r: vF(r)}; W1(x) = {r: xW(r)}; ,

    . PS = (X, V; P) WS0 = (X, R0, W0), WS1 = (Y1, R1; W1), ... ,

    WSk = (Yk, Rk; Wk), PSWSWSWS kk kki 1111 ... :}{

    69

  • k AS = (PS, WS1, ..., WSk; 1, ..., k), Yk Yk1 ... Y1 X i ( , )1 k ri Ri {ri1} Ri1, Wi(ri) Wi1({ri1}) &{ri1} - Wi1. , - WSi. k (X,V, R1, ..., Rk; F1, F2, ..., Fk) (X, V, R1, ..., Rk).

    , P, F, W PS, FS, WS , , - AS.

    - . , , , . , - xX rR, vV, xP(v) vF(r), xP(F(r)). , . .

    , . , - , - , . - - , . , x1 x2 X V, vV x1P(v) x2P(v), r, rR x1W(r) x2W(r). AS.

    AS , - PS, FS WS.

    1.2.

    S = (X, V, R; P, F, W) , vV P(v)= 2, rRW(r)= 2, rR F(r) V PS = (X, V). , PS WS S ,

    F WS = (X, R) PS = (X, V). F(r) , F -

    W. , F(r) - r, S (X, V, R; P, F).

    S = (X, V, R) S = (X, V, R ) , , X X, V V, R R P F, , - P W. S S , X X, R R, W. - , , PS PS . , X X, V V, P.

    70

  • 2. 2.1. .

    , - , :

    PS = (X, V) AS. () ( - ), () S(Vi) L(Vi) PS, .

    WS = (X, R) - PS = (X, V) ,

    +)},({}{1

    min )()(yxRir

    i

    jrFivi crlvS

    ,

    (1)

    {(x, y)} x, y WS. - i ri , l(ri) . , - WS, (1) .

    , NP- - .

    2.2 .

    - S = (X, V, R), s(ri) s(vi) -,

    )(:)(

    irFivirirs . ( )

    P, < .

    }1,0{,

    1)(:

    1min)()(

    =

    =+

    idicn

    kkrFivi

    iii

    n

    ii vsdrsc

    U.

    - . , , , .

    2.3. .

    , , [35]. , .

    . . AS = (X, V, R; P, F, W), X =

    (x1,..xn), V = (v1, v2,, vm), R = (r1, , rn), V(ri) ri. R R , :

    71

  • x X, I(x) R, URrRir

    i xrIrv

    = .)( min,)( I(x) ( -

    ) x, I(r) ( ) r. , . , , . - , .

    , 18 . , . . - ( ) - , .. .

    . X V- (R-),

    X V- (R-). - . , 18-. NP-.

    2.4. .

    AS = (X, V, R; P, F, W), Z(xi) xiX, l(vj) vj V, c(vj) vj.

    =

    )()()(

    krFjVjk vlrl ri .

    xi :

    +=})({1

    )()(),()},(),{(irFjv

    jXjx

    jjijii vcxZxxxxx ,

    {zi} , mi xi xj, = )(),( kji zlxx , zk(xi, xj), (xi, xj) xi xj.

    (x0) x0 - {(xi) | (xi, xj)} (x0, xj) ,

    ((xi), (xi, xj)) min (2) , x0

    . , |,|)()(||)( PSixiWSi

    PSocmiWS TxxTx ++

    )( iWS x WS =(X, R), PSocmT PS = (X,V), || PSocmT . ||

    PSixT ( ) PS

    || WSixT PS. WS

    PS , |||| PSixWSix TT = .

    .

    72

  • AS = (X, V, R, P, F, W), - x0 AS , (x0)-min , WS , .

    x0 AS, (2) - AS.

    2.5. .

    , , PS = (X, V), WS = (Y, R).

    (vij) PS(vij) vij V; (rij) WS(rij) rij R; (vij) () vij V; (rij) rij R; (S) S.

    S = (PS, WS; ), (S) k

    vV )()( xrr

    (S) S. - (S) .

    1. .. . 6. ,1981. . 2648. 2. .. . . . 2006. 3. . . . .: , 1978. 4. ., . , . .: , 1978. 5. .. . .: , 1987. , , . , 6, , 630090, , . (383) 330-96-43. E-mail: [email protected].

    73

  • . . , . .

    , ,

    . -

    F (x) =12 ||(Axb)+||2, ( ) - Ax b, Amn , b = (b1, . . . , bm) x = (x1, . . . , xn) . x+ = max(0, x) x p+ = (p+1 , . . . , p

    +n )

    p = (p1, . . . , pn). -

    [19]. , , (

    -

    ), ( -

    ), -

    (

    ), (

    - -

    , ) .

    ,

    .

    F (x) , - . , . . -

    H(x) = (2F/xixj)nn , - .

    . . [10, 11]

    F (x, x0) =1

    2(Ax b)TD0(Ax b) +

    2||x x0||2,

    x0 , , > 0 , D0 mm- Ax b x0, . . , i- 1, i- x0, 0 . - F (x) xk+1 = xk + kxk

    xk = (ATDkA+ E

    )1F (xk) = xk xk, E mm-, Dk mm- - Ax b xk, xk F (, xk)

    74

  • , k ( 1), - (0, 1) ,

    (, F (xk) = 1F (xk, xk))., -

    H(xk) = ATDkA+E. - H(xk)x = F (xk), xk, , , , -

    . -

    H(xk), . , , . . , ,

    .

    -

    , -

    ,

    -

    H = ATDA+ E (., , [1215]). , -

    H. - -

    A =

    A11 A1,n

    A22 A2,n.

    .

    .

    .

    .

    .

    An1,n1An1,n

    . , H - (, )

    H =

    H11 H1,n

    H22 H2,n.

    .

    .

    .

    .

    .

    Hn1,n1 Hn1,nHn,1 Hn,2 . . . Hn,n1 Hn,n

    , . , -

    [15],

    .

    .

    - -

    A =

    A11

    A22.

    .

    .

    An1,n1An,1 An,2 . . . An,n1

    .

    75

  • H = ATDA+E, , - , . . . -

    , H = ADTAT + E, , .

    ( ) -

    H = ATDA+ E.,

    .

    , -

    . , ,

    , ,

    .

    . -

    . -

    , , -

    -

    . , --

    ,

    , , -

    H = ADAT , H = ATDA . -

    . -

    ,

    . -

    .

    .

    -

    -

    -

    min (c, x) : Bixi = bi (i = 1, . . . , r), Ax = b0, x = [x1, . . . , xr] 0.

    c = [c1, . . . , cr] A = [A1, . . . , Ar], ci Rni ,Bi Ai mi ni- m0 ni- , m =

    ri=0mi, n =

    ri=1 ni. Ax = b0 .

    (x) =1

    2

    ri=1

    (||Bixi bi||2 + ||xi ||2)

    {xk Arg min{(x) : Ax = b0, (c, x) = k },k+1 = k + 2

    1k (x

    k), k = 0, 1, 2, . . . .

    76

  • -

    -

    , -

    k (k > 0). 0. , , , . .

    K xK . (x), - xs+1 = xs + sx

    s (s = 0, 1, . . . ), s = 1 ( -

    ),

    xs = xs xs. xs

    (x, xs) =1

    2

    ri=1

    (||Bixi bi||2 + xTi D(i)s xi)+ 2 ||x xs||2 - Ax = b0, (c, x) = k -

    (BTi Bi +D(i)s + Enini) x

    si + A

    Ti y + sc = B

    Ti bi + x

    si (i = 1, . . . , r),

    A1 xs1 + . . . + Ar x

    sr = b0,

    (c1, xs1) + + (cr, xsr) = k, D

    (i)s = diag( sign((xsi )+1 ), . . . , sign((xsi )ni+) ), E - , s , > 0 - ( ).

    -

    . ,

    . -

    y = H10[ ri=1

    AiH1i

    (pi sci

    ) b0

    ], xsi = H

    1i (pi ATi y sci), i = 1, . . . , r,

    H0 =ri=1

    AiH1i A

    Ti , Hi = B

    Ti Bi +D

    (i)s + Enini , pi = B

    Ti bi + x

    si , i = 1, . . . , r,

    s =

    k + (ri=1

    cTi H1i A

    Ti )H

    10 (

    ri=1

    AiH1i pi b0)

    ri=1

    cTi H1i pi

    (ri=1

    cTi H1i A

    Ti )H

    10 (

    ri=1

    AiH1i ci)

    ri=1

    cTi H1i ci

    .

    , r Hi nini H0 m0m0, Hi, i = 1, . . . , r, . [16].

    , 07-01-00399.

    77

  • 1. Eremin I.I. Theory of Linear Optimization. Inverse and Ill-Posed Problems Series. VSP.

    Utrecht, Boston, Keln, Tokyo, 2002.

    2. .., .. -

    . M.: , 1979.

    3. .. -

    . .: , 1982.

    4. .. . .: . .

    . .-.., 1988.

    5. .., .. . .: ,

    2003.

    6. .. ,

    .- .: (), 1979.

    7. .. // , 1974,

    .14, 4, 10521058.

    8. ..

    // . 1977, N 1, .5-15.

    9. .., .. -

    // . . 1972.

    .8. .5. .740751.

    10. Mangasarian O.L. A nite Newton method for classication // Optimizat. Meth.

    Software. 2002. Vol.17, p.913930.

    11. Kanzow C., Qi H., Qi L. On the minimum norm solution of linear program // J.

    Optimizat. Theory and Appl. 2003. Vol.116. p. 333345.

    12. .., .., .

    // . -

    . . . , 2004, 44, 9, . 15641573.

    13. ., . : . . .: , 1999.

    14. .

    / . . .: , 1991.

    15. Gondzio J., Sarkissian R. Parallel interior-point solver for structured linear programs

    // Math. Progam., Ser. A. 2003, Vol. 96, p. 561584.

    16. .. -

    // . . -

    . . . 2007. 47, N2, .206-221.

    , ,

    . . , 16, , 6200219, ,

    . (8-343-3)75-34-23, (8-343-3)74-25-81, e-mail:[email protected]

    ,

    78

  • . .

    [1], -

    . -

    , -

    . -

    . , , [2],

    c, x+ d, y max(x,y)

    ,

    x > 0, Ax+By 6 b,y Y(x) , Sol(2),

    (1)c1, x+ d1, y max

    y,

    y > 0, A1x+B1y 6 b1,

    }(2)

    - (2) -

    ( )

    d1 + vB1, y v, b1 A1xB1y = 0,x, y, v > 0.

    , , d.c.

    , .. .

    , -

    , x,

    x > 0, Mx+ q > 0,x,Mx+ q = 0.

    }(3)

    , ( )

    -

    , , , -

    , ..

    , P : X IRm, X IRn, x X,

    P (x), x x > 0 x X. (4) (3) (4) P (x) = Mx + q, X = IRm+ . (3) (4).

    , (

    ) -

    ( ).

    , - ,

    , ,

    , , .

    79

  • , , -

    f0 min, fi(x) = 0, i = 1, . . . ,m, (5)

    fi(x), i = 1, . . . ,m, (..

    0f0(x) +mi=1

    ifi(x) = 0; (6)

    i IR, i = 1, . . . ,m, 0 > 0,mi=1

    |i| + 0 > 0), . , , (1)-(2), (3) (4)

    , (6)

    .

    ,

    ()

    (x) = f0(x) + (x) minx(7)

    (x) = 0(f1(x), . . . , fm(x)) , , :

    1(x) =mi=1

    i|fi(x)|, 2(x) =mi=1

    i[fi(x)]2, p(x) =

    mi=1

    i[fi(x)]p, 1 < p < +.

    , fi(x), i = 1, . . . ,m - (x), . (7) (

    ) ( -

    ) . ,

    (5) ,

    d. c. -

    :

    (P) f(x) min, x S,F (x) , g(x) h(x) = 0,

    }(8)

    f, g, h , S IRn. , (P)(8) d. c. :

    f(x) min, x S,F (x) , g(x) h(x) > 0.

    }(9)

    (9) [6], -

    [4],[5],

    [3],[5]. , -

    (9) .

    (9) -

    (8).

    80

  • . , (P)(8) :

    v S : F (v) < 0, (10)

    (H) y S : F (y) = 0 . . g(y) = h(y), p = p(y) S :h(p) h(y) < g(y), p y.

    }(11)

    [3],[4] -

    ().

    1 (e )[3],[4]. -

    (10) (H)(11). ,

    (E) (y, ) : g(y) = , y S,

    h(y) 6 6 sup(h, S),h(x) > g(y), x y,x S, f(x) 6 f(z);

    (12) z (P)(8). # (E)(12) (P)(8) - , (9). -

    , , (10):

    v S : f(v) < f(z), F (v) < 0. (13) 2 ( ). (P)-(8) (13).

    , z Sol(P),

    (E1)(y, ) : g(y) = , y S,h(y) > g(y), x yx S, f(x) 6 f(z).

    (14). (E1)-(14) ,

    (y, ) : g(y) = , y S, u S, f(u) 6 f(z),h(u) < g(y), u y. g()

    0 < h(u) + g(u) g(y) = F (u),

    u S, f(u) 6 f(z), F (u) > 0. (13), ]0, 1[: F (x()) = 0,

    x() = u+ (1 )v S. , f(),

    f(x()) 6 f(u) + (1 )f(v) < f(z),

    81

  • z. # (E)-(12) (E1)-(14) (1)-(2) (3) - ,

    .

    , (1)-

    (2),

    .

    1. .. . .: , 1975.

    2. Dempe S. Foundations of bilievel programming. Dordrecht/ Boston/London: Kluwer

    Academic Publishers, 2002.

    3. .. , : ,

    2003.

    4. .. d.c. .//

    . . . , 2001, .41, 12, . 1833-1843.

    5. .. d.c. -

    .// . . . , 2005, .45, 3, . 435-447.

    6. .., .. -

    // . . . . 2007, .47, 3, c. 397-

    413.

    ,

    . 134, , 644033, , . (3952)511398.

    E-mail: [email protected]

    82

  • . .

    , , .

    F. X Rn. x X , X = . , F. -

    , , X

    X =

    D, (1)

    = {x Rn : < xj xj xj

  • , .. -

    , , ,

    , MAPLE -

    [3].

    [6].

    . g(x, y) = sin(xy)+0.1(x 1)2+0.2y2, X = = {x R2 : 3 xj 3, j = 1, 2} y = (0, 0). y -

    (x, y) = min

    (0,(x y)

    2

    2

    )+ 0.1 0.2x.

    (2)-(3) .

    D

    = B = {x Rn : 0 xj 1, j = 1, . . . , n},D = DB = {x Rn : hB(x) = xT (x e) = 0}, e = (1, . . . , 1)T . XB = B DB B. D

    D = DI = {x Rn : hI(x) =nj=1

    | sin(pixj)| = 0},

    x DI x Zn, Zn n- , .. DI = Zn. , x

    xj Zj = {zj1, . . . , zjkj}, j = 1, . . . , n. (4) (4).

    j(xj) =

    kjs=1

    (x zjs).

    (4) x DZ ,

    DZ = {x Rn : hZ(x) =nj=1

    |j(xj)| = 0}.

    hB(x) , hI(x) hZ(x) - .

    D (3)

    D = {x Rn : F (x) 0},F (x) = max{g1(x), . . . , gm(x), |h1(x)|, . . . , |hl(x)|}.

    84

  • F (x) - F (x, y). x0 - . F (x0) 0, x0 X .

    x0 6 X (5) F (x0) > 0. F (x

    0, x0) = F (x0),

    F (x0, x0) > 0. (6)

    , F (x, x0) F (x) x X.

    F (x, x0) 0 x X. (7) (5)-(7) , F (x, x

    0) = 0 x0 X. , . -

    ,

    2

    F (x, y). , x0 . -

    U(x0) = {x : F (x, x0) > 0}, X. C0 - x0, . C0 F (x, x

    0) = 0

    (p0)Tx = s0. (8)

    U(x0) (8) , ..

    (p0)Tx0 > s0 (p0)Tx s0 x X.

    P 0 = {x : (p0)Tx s0}., x0 6 P 0 P 0 X. () x1 P 0 x1, P 1 ,

    P 0 P 1 X.

    P 0 P 1 . . . P k X xk, k = 0, 1, . . . , k. , P k = k, - X = , limk xk = x X. ,

    ,

    2

    .. , .

    85

  • . Ck

    Cki ,

    Ck =i

    Cki int(Ckj)

    int(Cki) = i 6= j.

    , -

    , ,

    . -

    . -

    X. -

    , -, ,

    (1)-(3). . -

    , [5].

    .

    06-01-00465-

    1. .. , .. . . // -

    . 2. 2004. . 44. N9. . 45-68.

    2. .. . -

    . // XIII - " -

    ", .1 " ", ,

    . 2005. . 621-626.

    3. .. . , -

    . // .: , , . 1998. . 73-100.

    4. .. . .

    // . 2004. . 44. N9.

    . 1552-1563.

    5. V.P. Bulatov, O.V. Khamisov. The branch and bound method with cuts in En+1

    for solving concave programming problem. // Lecture Notes in Control and Computer

    Sciences, 180, Springer-Verlag. 1991. P. 273-282.

    6. G.P. McCormik. Nonlinear programming: Theory, Algorithms and Applications. //

    John Wiley and Sons. New York. 1988. 267 p.

    , , . 130, , 664033,

    , . (3952) 42-84-39, (3952) 42-67-96, E-mail: [email protected]

    86

  • ,

    . .

    [1,2] MASC

    A,B Qn. , , NP - Apx( P 6= NP ). - MASC.

    , ,

    . , -

    , (..

    ).

    PC

    , -

    [3]. , [4], -

    -

    .

    1. Q = (f1, . . . , fq), fi(x) =Ti x i , A,B Rn,

    |{i Nq | fi(a) > 0}| > q2

    (a A),

    |{i Nq | fi(b) < 0}| > q2

    (b B).

    q () Q.

    (MASC).

    A = {a1, . . . , am1} B = {b1, . . . , bm2}, A,B Qn. Q , - A B.

    1 ([2]). MASC NP-.

    MASC NP -

    A B {z {0, 1, 2}n : |z| 2}.

    2 ([2]). MASC Apx ( P 6= NP ).

    87

  • MASC.

    3.

    NP 6 TIME(2poly(logn)), MASC -

    O(log log logm).

    , -

    . [5], n = 1 .

    , n = 2 (, , n > 1) NP -.

    2. L = {l1, . . . , ls}, lj = {x R2 | cTj x = dj}, P = {p1, . . . , pk} R2, p P l = l(p) L , p l. , , ,

    P,A B, , , - . ,

    ,

    .

    . 1: PC PASC

    (PC).

    P = {p1, . . . , pk} Z2 s N. L P s?

    88

  • (PASC).

    A = {a1, . . . , am1} B = {b1, . . . , bm2}, A,B Q2, t N. Q, A B t ?

    [3], PC NP -. PASC ASC [2], .

    , PASC ( ASC) NP. PC PASC,

    NP - . PC P = {p1, . . . , pk} Z2 s N. = max{|pi| : i Nk} = 16(2+1)+1 . - , || = 1 , {i, j} Nk [pi , pi + ] [pj, pj+] . PC PASC : A = P, B = (P ) (P + ) t = 2s+ 1 (. 1). ,

    , PC.

    , -

    PC PASC

    . , P - s , A B , 2s+ 1.

    4. P = {p1, . . . , pk} Z2 s , A = P B = (P ) (P +) 2s+ 1 .

    . 2:

    89

  • ,

    ,

    P , A B, . 2.

    1. PASC NP-. ASC

    n > 1 NP-.

    2. MASC n > 1 NP-.

    , -6768.2006.1 -

    5595.2006.1 , 07-07-00168.

    1. ...

    // , 2006, 406, 6, . 742745.

    2. ... -

    . //

    . 2006, 1, . 3443.

    3. N.Megiddo, A.Tamir. On the complexity of locating linear facilities in the plane //

    Operations research letters. 1982, vol. 1, no. 5, p. 194197.

    4. N.Megiddo. On the complexity of polyhedral separability // Discrete and Computational

    Geometry. 1988, 3, p. 325337.

    5. ... // -

    . 1971. 3. . 140146.

    , ,

    . . , 16, , 620219, ,

    . (343) 375-35-05, (343) 374-25-81, E-mail:[email protected]

    90

  • f.

    . .

    A = {a1, ..., an} aj Rd, ( [A]) d- . - A TA = {S1, S2, ..., St} Si A, [Si] d- , |Si| = d + 1,

    ti=1

    [Si] = [A] i 6= k[Si Sk] = [Si] [Sk]. j = 0, 1, 2, ..., d (j + 1)- F A j- TA, i , F Si. fj(TA) j- f(, TA) =

    d+1j=0

    fj1(TA)j, f1(TA) = 1.

    TA A - f(, TA) F (d, n). ( -

    , ) f() ( F (d, n)).

    , . . . -

    , . , . 23, . , 603950, , . (8-8312) 65-78-81,

    E-mail:[email protected]

    91

  • CONTINUOUS COVERING PROBLEMS

    P. Hansen

    Covering problems are frequently encountered in Operations Research, Location Theory,

    Telecommunications and Geometry. The most studied are the discrete ones, such as the

    p-center problem. However, continuous problems are of interest also. They are of two

    types:

    (i) discrete-continuous ones, in which a discrete set of demand points is given, together

    with a continuous set wherein facilities are to be located, the objective being to minimize

    the maximum distance from a demand point to its closest facility;

    (ii) fully continuous ones which dier from the former only in that the set of demand

    points is continuous; this last category comprizes well-known geometric problems such as

    covering disks, squares or tringles by a minimum number of disks of given radius (or with

    a given number n of disks with minimum radius).

    We review work on these problems and provide new heuristic and exact algorithms for

    both of them.

    Pierre Hansen,

    GERAD and Department of Quantitative Methods in Management, HECMontreal, Canada,

    phone: (1-514) 340-6052, fax: (1-514) 340-5665. E-mail: [email protected]

    92

  • THE VARIABLE NEIGHBORHOOD SIMPLEX SEARCH

    FOR CONTINUOUS OPTIMIZATION

    Q. Zhao, D. Urosevic and N. Mladenovic

    We rst suggest a modied version of the well-known Nelder-Mead (or simplex) method,

    originally designed for solving continuous convex minimization problems. Then we propose

    a natural and simple extension that allows us to solve non convex nor concave problems

    as well. It ts into the variable neighborhood search scheme. Extensive computational

    analysis shows the capability of our method. It appears that, in solving convex problems,

    our modied simplex outperforms in average the original version as well as some other

    recent modications. In solving unconstrained global optimization, it is comparable with

    the state-of-the-art heuristics, but easier to implement and more user-friendly.

    Nenad Mladenovic

    School of Mathematics, University of Birmingham Edgbaston, Birmingham B15 2TT,

    United Kingdom, e-mail: [email protected].

    93

  • . . , . . , . .

    n- , - . , -

    .

    .

    i j cij(xij, xji) ( xij, xji , - i j (i, j) (j, i)) xij . , , -

    , , ,

    .

    [1] -

    cij(xij, xji) = axij + bxji, .

    :

    1. cij(xij, xji) = aixij + ajxji;

    2. cij(xij, xji) = aij(xij + xji);

    3. cij(xij, xji) = aixij + bijxji.

    1 2 -

    O(n3).

    1. .., .. //

    . 2005. . 8, 3(23). . 5868.

    , () -

    - , . , .55-, .130,

    650055, , , . (8-3842) 25-33-34, (8-3842) 25-07-21. E-mail:

    [email protected]

    , . .. , . -

    , 4, 630090, , , . (8-383) 333-37-88, (8-383) 333-25-98.

    E-mail: [email protected]

    , , . -

    , 2, 630090, , . E-mail: [email protected]

    94

  • ,

    . . , . . , . .

    :

    x = argmin{(x) : x R}, (1)

    (x) x En, R En . : R , - x (1), {Sk} , |Sk| 0 x Sk k. Sk+1 xk Sk, -. .

    [1, 2], :

    1.

    (Sk)

    (Sk1)(

    n1n1 1

    )n11(

    n

    n+ 1

    )n< 1 (2) Sk, n1 . , R - Ax b A. , -

    1 n1 n, (2) :|Sk||Sk1|

    1

    n

    [ nn12

    (n1

    n1 1)n11

    (

    n

    n+ 1

    )n+1

    2

    ] 1 lnn

    n.

    2.

    |Sk||Sk1|

    (k

    k 1)k1

    (

    k

    k + 1

    )k< 1 Sk, k (1 k n) Sk, . , - , n , :

    |Sk||Sk1|

    1

    n

    [ nk2

    (k

    k 1)k1

    (

    k

    k + 1

    )k+1

    2

    ] 1 0.79

    n.

    3. xk Sk {Sk} -, :

    |Sk||Sk1|

    (1 n

    n+ 1

    )< 1.

    , -

    .

    06-01-00465

    95

  • 1. .., .. -

    . // . : , 1982.

    2. .., ..

    // . 1984. . 27, 3. . 348385.

    , ,

    . .. 664033, . , .

    , 130 . 8(3952)42-84-40, e-mail: [email protected]

    ,

    , 664004, . ,

    . , 14

    96

  • . .

    () M Rn : Rn Rn M , x / M |(x)M | < |xM |, |xM | =infyM

    |x y|. ,

    [1,2]. , ,

    , .

    1 - , M Rn, x0 Rn , xn = (xn1). {xn} - M . -, {xn} M , M ., M [3], - r > 0 : M + rB Unp(M), Unp(M) - x Rn - piM(x) M , B - . C2- [3] . 2 - , - M Rn, x0 Rn, xn+1 = (xn) n=0

    [xn, xn+1] M = . (*) N, > 0 : x {xn} , n > N cos ((x) x, piM(x) x) > . xn M. M Rn - , M - (*).

    , -5595.2006.1.

    1. .., .. -

    . . : , 2005.

    2. .., .. -

    , .: "", 1979.

    3. Federer H., Curvature Measures // Trans. Amer. Math. Soc. 93, 3 (1959), 418-493.

    , , -

    4, , 620083, , . 8-902-253-39-40, e-mail: [email protected]

    97

  • ,

    . . , . .

    , -

    ( ), -

    .

    .. , (

    10.1 [1]). -

    ,

    . ,

    . ,

    .

    .

    ,

    . , -

    , -

    .

    , ,

    , -

    . -