min f(x)
h(x) = 0
h(x) = [h1(x)h2(x)...hm(x)] = 0 { h1(x) = 0h2(x) = 0...hm(x) = 0
1)x(0)punto iniziale rispetto alquale approssimato i vincolih(x) con i polinomio di Taylorarrestato al primo ordine
hi(x) = hi(x(0)) + Vhi(x(0))T(x - x(0))
x - x(0) = dx
hi(x) = hi(x(0)) + Vhi(x(0))T dx = 0
h(x) = h(x(0)) + Jh(x(0)) dx = 0
dx = x - x(0)
2) APPROSSIMIAMO LA f(x) CON IL POLINOMIO DI TAYLOR ARRESTATO AL SECONDO ORDINE ATTORNO AL PUNTO x(0)
f(x) = f(x(0)) + [del]f(x(0))T(x - x(0)) +
1/2(x - x(0))T Hf(x(0))(x - x(0)).
min f(x(0)) + [del]f(x(0))T(x - x(0)) +
1/2(x - x(0))T Hf(x(0))(x - x(0))
h(x) = h(x(0)) + J h(x(0)) dx
vettore vettore matrice vettore vettore
vettore
min f(x(0)) + [del]f(x(0))T dx +
1/2 dxT Hf(x(0)) dx +
Jh(x(0)) dx = 0
L(dx, λ) = f(x(0)) + ∇f(x(0))T dx +
(1/2) dxT Hf(x(0)) dx + λ (h(x(0)) +
Th(x(0)) dx)
{∇dxL(dx,λ) = 0 => ∇f(x(0)) + Hf(x(0)) dx + λT Th(x(0)) = 0
∇λL(dx,λ) = 0 => h(x(0)) + Th(x(0)) dx = 0
x(0) -> dx = x - x(0)
λ(0) -> dλ = λ - λ(0)
λ = dλ + λ(0)
∇dxL(x, λ) = ∇f(x(0)) + Hf(x(0)) dx + ddT Jh(x(0)) + λ(0)T Jh(x(0)) = 0
h(x(0)) + Jh(x(0)) dx = 0
Hf(x(0)) dx + J(h(x(0)))Tdλ = -∇f(x(0)) - λ(0)T Jh(x(0))
Jh(x(0)) dx = -h(x(0))
-(∇f(x(0)) + λ(0)T Jh(x(0)))
∇L(x(0), λ)dx
L(x, λ) = f(x) + λTh(x)
∇dxL(x, λ) = ∇f(x) + λTJh(x)
DelL(x(0),l(0)) = Df(x(0)) + J h(x(0))
+ J h(x(0))
x
∫ H f(x(0)) d x + J h(x(0))T d λ = - ∇L(x(0),l(0))dx
J h(x(0)) d x = - h(x(0))
[ H f(x(0)) J h(x(0))T ] [ d x ] = [ - ∇L(x(0),l(0))dx ]
[ J h(x(0)) 0 ] [ d λ ] = [ - h(x(0)) ]
∫ H f(x(0)) d x + J h(x(0))T d y = - ∇d xL(x(0),l(0))
J h(x(0)) d x = - h(x(0))
(dx, dL) = M-1 w
[dx] = x - x(0)
[dL] = λ - λ(0)
x(1) = dx + x(0)
λ(1) = dλ + λ(0)
x(0) = arbitrario
mimim x14 + x24
x12 + x22 = 2
h(x) = x12 + x22 - 2 = 0
x(0) = (1, 1)
λ(0) = 1
∇f(x(0)) = ∇f(x)|x=x(0) = [4x13, 4x23]x=x(0) = [4, 4]
Hf(x(0)) = [12x12 0, 0 12x22]x=x(0) = [12 0, 0 12]
∇L(x(0), λ(0)) = ∇f(x(0)) + λ(0)Jh(x(0)) = [4, 4] + 1 [2, 2]
h(x(0)) = 0
Jh(x(0))Tλ(0) = [6, 6]
Jh(x0) = Vh(x0)T = [2x1, 2x2] |x = x0 = [22
[∇f(x0)T]
M =
[ 12 0 2 ]
[ 0 12 -2 ]
[ 2 2 0 ]
det M = -48-48 = -86
w =
[-6]
[-6]
[-1]
M-1 = -1/96 [ -4 4 -24 ]T =
[ 4 -4 -24 ]
[-24 -24 144 ]
= -1/96 [ -4 4 -24 ] = [ 1/24 -1/24 1/4 ]
[ 4 -4 -24 ]
[-24 -24 144 ]
[ -1/24 1/24 1/4 ]
[ 1/4 1/4 3/2 ]
M-1w =
[ 1/24 -1/24 1/4 ] [ -6 ] =
[ -1/24 1/24 1/4 ] [ -6 ]
[ 1/4 1/4 3/2 ] [ 0 ]
[d(x)8(x)] - [ -1/8 + 1/4 -1/3; 1/4 -1/8 -1/3; -1/2 -3/2 0 ] = [ -1/4; 1/8; -3 ]
[x(1)1x(1)2x(1)3] = [ -1/4; -1/6; -3/2 ] + [1; 1; 1] = [ 3/4; 3/4; -2 ]