7 篇文章 2 订阅

# 一、问题描述

min ⁡ f 0 ( x ) \min f_0(x)

s . t . f i ( x ) ≤ 0 , i = 1 , 2 , ⋅ ⋅ ⋅ , m s.t.\quad f_i(x)\le0,\quad i = 1, 2, \cdot\cdot\cdot, m

h i ( x ) = 0 , i = 1 , 2 , ⋅ ⋅ ⋅ , p \qquad h_i(x)=0,\quad i = 1, 2, \cdot\cdot\cdot, p

L ( x , λ , ν ) = f 0 ( x ) + ∑ i = 1 m λ ( i ) f i ( x ) + ∑ i = 1 p ν ( i ) h i ( x ) ( 公 式 1 ) L(x,\lambda,\nu)=f_0(x)+\sum_{i=1}^{m}\lambda_{(i)}f_i(x)+\sum_{i=1}^{p}\nu_{(i)}h_i(x) \qquad (公式1)

g ( λ , ν ) = inf ⁡ x ∈ D L ( x , λ , ν ) = inf ⁡ x ∈ D ( f 0 ( x ) + ∑ i = 1 m λ ( i ) f i ( x ) + ∑ i = 1 p ν ( i ) h i ( x ) ) ( 公 式 2 ) g(\lambda,\nu)=\inf\limits_{x\in D}L(x,\lambda,\nu)=\inf\limits_{x\in D}(f_0(x)+\sum_{i=1}^{m}\lambda_{(i)}f_i(x)+\sum_{i=1}^{p}\nu_{(i)}h_i(x)) \qquad (公式2)

# 二、证明

g ( θ λ 1 + ( 1 − θ ) λ 2 , θ ν 1 + ( 1 − θ ) ν 2 ) ≥ θ g ( λ 1 , ν 1 ) + ( 1 − θ ) g ( λ 2 , ν 2 ) , θ ∈ R ( 公 式 3 ) g(\theta\lambda_1+(1-\theta)\lambda_2,\theta\nu_1+(1-\theta)\nu_2)\ge\theta g(\lambda_1,\nu_1)+(1-\theta)g(\lambda_2,\nu_2),\quad\theta\in R\quad(公式3)

g ( λ , ν ) = m i n { L ( x 1 , λ , ν ) , L ( x 2 , λ , ν ) , ⋅ ⋅ ⋅ , L ( x n , λ , ν ) } , n = + ∞ ( 公 式 4 ) g(\lambda,\nu)=min\{L(x_1,\lambda,\nu),L(x_2,\lambda,\nu),\cdot\cdot\cdot,L(x_n,\lambda,\nu)\},\quad n=+\infty\qquad(公式4)

g ( θ γ 1 + ( 1 − θ ) γ 2 ) = m i n { L ( x 1 , θ γ 1 + ( 1 − θ ) γ 2 ) , L ( x 2 , θ γ 1 + ( 1 − θ ) γ 2 ) , ⋅ ⋅ ⋅ , L ( x n , θ γ 1 + ( 1 − θ ) γ 2 ) } ( 公 式 5 ) g(\theta\gamma_1+(1-\theta)\gamma_2)=min\{L(x_1,\theta\gamma_1+(1-\theta)\gamma_2),L(x_2,\theta\gamma_1+(1-\theta)\gamma_2),\cdot \cdot\cdot,L(x_n,\theta\gamma_1+(1-\theta)\gamma_2)\}\qquad(公式5)

≥ m i n { θ L ( x 1 , γ 1 ) + ( 1 − θ ) L ( x 1 , γ 2 ) , θ L ( x 2 , γ 1 ) + ( 1 − θ ) L ( x 2 , γ 2 ) , ⋅ ⋅ ⋅ , θ L ( x n , γ 1 ) + ( 1 − θ ) L ( x n , γ 2 ) } ( 公 式 6 ) \ge min\{\theta L(x_1,\gamma_1)+(1-\theta)L(x_1,\gamma_2),\theta L(x_2,\gamma_1)+(1-\theta)L(x_2,\gamma_2),\cdot\cdot\cdot,\theta L(x_n,\gamma_1)+(1-\theta)L(x_n,\gamma_2)\}\quad (公式6)

≥ θ m i n { L ( x 1 , γ 1 ) , L ( x 2 , γ 1 ) , ⋅ ⋅ ⋅ , L ( x n , γ 1 ) } + ( 1 − θ ) m i n { L ( x 1 , γ 2 ) , L ( x 2 , γ 2 ) , ⋅ ⋅ ⋅ , L ( x n , γ 2 ) } ( 公 式 7 ) \ge\theta min\{L(x_1,\gamma_1),L(x_2,\gamma_1),\cdot\cdot\cdot,L(x_n,\gamma_1)\}+(1-\theta)min\{L(x_1,\gamma_2),L(x_2,\gamma_2),\cdot\cdot\cdot,L(x_n,\gamma_2)\}\quad (公式7)

= θ g ( γ 1 ) + ( 1 − θ ) g ( γ 2 ) ( 公 式 8 ) =\theta g(\gamma_1)+(1-\theta)g(\gamma_2)\quad (公式8)

# 三、解释证明过程

（公式5）到（公式6）是因为 L ( x i , γ ) L(x_i, \gamma) 中的 x x 的值已固定，所以 f i ( x ) , i = 0 , 1 , 2 , ⋅ ⋅ ⋅ , m f_i(x), i = 0, 1, 2,\cdot\cdot\cdot, m h i ( x ) , i = 0 , 1 , 2 , ⋅ ⋅ ⋅ , p h_i(x), i = 0, 1, 2,\cdot\cdot\cdot, p 都应该看做常数，所以此时的 L ( x i , γ ) L(x_i, \gamma) γ \gamma 的仿射函数，而仿射函数是既凸且凹的，对（公式5）右边中的每一个拉格朗日函数都运用其凹性，就可以得到（公式6）.

a = { a 1 , a 2 , ⋅ ⋅ ⋅ , a n } b = { b 1 , b 2 , ⋅ ⋅ ⋅ , b n } a=\{a_1,a_2,\cdot\cdot\cdot,a_n\}\\b=\{b_1,b_2,\cdot\cdot\cdot,b_n\}

m i n { a i + b j } ≥ m i n { a } + m i n { b } , i , j ∈ N + min\{a_i+b_j\}\ge min\{a\}+min\{b\},\quad i,j\in N^+

（公式7）到（公式8）由公式4可得。

• 48
点赞
• 101
收藏
觉得还不错? 一键收藏
• 打赏
• 17
评论
03-21 1228
05-19 2万+
04-05 1288
02-09 2112
01-04
11-18 9794
07-09 3886

### “相关推荐”对你有帮助么？

• 非常没帮助
• 没帮助
• 一般
• 有帮助
• 非常有帮助

u014540876

¥2 ¥4 ¥6 ¥10 ¥20

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、C币套餐、付费专栏及课程。