基本定义和概念
广义定义:目标函数凸,约束凸
一般优化问题:
min
f
0
(
x
)
,
s
.
t
.
f
i
(
x
)
≤
0
,
i
=
1
,
⋯
,
M
,
h
i
(
x
)
=
0
,
i
=
1
,
⋯
,
P
\min f_0(x),s.t.f_i(x)\leq 0,i=1,\cdots,M,h_i(x)=0,i=1,\cdots,P
minf0(x),s.t.fi(x)≤0,i=1,⋯,M,hi(x)=0,i=1,⋯,P
优化变量(optimization variable):
x
:
R
n
x:R^n
x:Rn,
目标函数\损失函数(objective function cost function)(max的话叫做效用函数 utility function): f 0 : R n → R f_0:R^n\rightarrow R f0:Rn→R
不等式约束(inequality constraint): f i : R n → R f_i:R^n\rightarrow R fi:Rn→R
等式约束(equality constraint): h i : R n → R h_i:R^n\rightarrow R hi:Rn→R
Notes:如果 M = P = 0 M=P=0 M=P=0,叫做无约束优化问题
优化问题的域(dominant): D = ⋂ i = 0 M d o m ( f i ) ∩ ⋂ i = 0 P d o m ( h i ) D=\bigcap_{i=0}^M dom(f_i)\cap \bigcap_{i=0}^P dom(h_i) D=⋂i=0Mdom(fi)∩⋂i=0Pdom(hi)
可行解集(feasible set): X f = { x ∣ x ∈ D , f i ( x ) ≤ 0 , h i ( x ) = 0 } X_f=\{x|x\in D,f_i(x)\leq 0,h_i(x)=0\} Xf={x∣x∈D,fi(x)≤0,hi(x)=0}
最优值(optimal value): p ∗ = inf { f 0 ( x ) ∣ x ∈ X f } p^*=\inf\{f_0(x)|x\in X_f\} p∗=inf{f0(x)∣x∈Xf}
Notes:如果 X f X_f Xf是空集, p ∗ = + ∞ p^*=+\infty p∗=+∞
最优解集(optimal set): X o p t = { x ∗ ∣ x ∗ ∈ X f , f 0 ( x ∗ ) = p ∗ } X_{opt}=\{x^*|x^*\in X_f,f_0(x^*)=p^*\} Xopt={x∗∣x∗∈Xf,f0(x∗)=p∗}
ϵ \epsilon ϵ次优解集( ϵ \epsilon ϵ-suboptimal set): X ϵ = { x ∗ ∣ x ∗ ∈ X f , f 0 ( x ∗ ) ≤ p ∗ + ϵ } X_{\epsilon}=\{x^*|x^*\in X_f,f_0(x^*)\leq p^*+\epsilon\} Xϵ={x∗∣x∗∈Xf,f0(x∗)≤p∗+ϵ},真正需要的不一定是最优解,通常满足一定的条件和误差就够了。(satisfy suffice->satisfice solution)
局部最优解(locally optimal): f 0 ( x ) = inf { f 0 ( z ) ∣ f i ( z ) ≤ 0 , h i ( z ) > 0 , ∥ z − x ∥ ≤ R } f_0(x)=\inf \{f_0(z)|f_i(z)\leq 0,h_i(z)>0,\|z-x\|\leq R\} f0(x)=inf{f0(z)∣fi(z)≤0,hi(z)>0,∥z−x∥≤R},存在 R R R使得 x x x局部最优
局部最优解集(locally optimal set):
X
l
o
c
X_{loc}
Xloc
活动约束(active):如果对于可行解集中的 x x x,存在 f i ( x ) = 0 f_i(x)=0 fi(x)=0,那么此项约束 f i ( x ) ≤ 0 f_i(x)\leq 0 fi(x)≤0称为活动约束。
不活动约束(inactive):如果对于可行解集中的 x x x,都有 f i ( x ) < 0 f_i(x)<0 fi(x)<0,即不等式约束严格成立,那么此项约束 f i ( x ) ≤ 0 f_i(x)\leq 0 fi(x)≤0称为不活动约束。