Forgot password?
 Register account
View 323|Reply 10

[不等式] 3-variables inequality

[Copy link]

7

Threads

2

Posts

69

Credits

Credits
69

Show all posts

jhsinutopia Posted 2024-4-8 14:28 |Read mode
$0\le\alpha,\beta,\gamma\le1$ and $4\alpha+\beta+3\gamma=\dfrac{9}{2}$.
$\min((\alpha-1)\beta+(\beta-1)\gamma+(\gamma-1)\alpha)=?$

686

Threads

110K

Posts

910K

Credits

Credits
91229
QQ

Show all posts

kuing Posted 2024-4-8 16:22
消去任意一个变量,目标函数都是另外两个变量的开口向下的二次函数,所以必在端点处取最小值。
而变量的可行域在空间中是一个多边形,那只需要计算出它的所有顶点坐标,代入目标函数看哪个最小即可。

3159

Threads

7941

Posts

610K

Credits

Credits
63770
QQ

Show all posts

hbghlyj Posted 2024-4-9 00:04
jhsinutopia 发表于 2024-4-8 06:28
$0\le\alpha,\beta,\gamma\le1$ and $4\alpha+\beta+3\gamma=\dfrac{9}{2}$.
$\min((\alpha-1)\beta+(\beta-1)\gamma+(\gamma-1)\alpha)=?$
NMinimize[{(α - 1) β + (β - 1) γ + (γ - 1) α,
  0 <= α <= 1 && 0 <= β <= 1 && 0 <= γ <= 1 &&
   4 α + β + 3 γ == 9/2}, {α, β, γ}]

{-1., {α -> 1., β -> 3.2333969609422432*^-9, γ -> 0.166667}}

3159

Threads

7941

Posts

610K

Credits

Credits
63770
QQ

Show all posts

hbghlyj Posted 2024-4-9 04:40
Last edited by hbghlyj 2024-4-9 08:23一般情况见forum.php?mod=redirect&goto=findpost& … d=7758&pid=59268
$Q(y)=\frac{1}{2} y^{\top} C^{-1} y-b^{\top} y$
$Q(y)$有全局最小值$\iff C$正定
而$Q(y)$限制在$A^\top y=f$上有最小值$\iff A^\top CA$正定


$C$正定,则$A^\top CA$正定
但$A^\top CA$正定不能推出$C$正定(当$C$非正定时,$A^\top CA$也可能正定,例如此题$C^{-1}=\pmatrix{0&1&1\\1&0&1\\1&1&0},A=\pmatrix{4\\1\\3}$,则$C=\frac12\pmatrix{-1&1&1\\1&-1&1\\1&1&-1}$非正定,而$A^\top CA=\pmatrix{6}$正定,故$Q(y)$无全局最小值但限制在$A^\top y=f$上有最小值)

7

Threads

2

Posts

69

Credits

Credits
69

Show all posts

 Author| jhsinutopia Posted 2024-4-9 07:11
Last edited by jhsinutopia 2024-4-9 07:49its typo, sorry! my mistake. I want
$\min((1-\alpha)\beta+(1-\beta)\gamma+(1-\gamma)\alpha)=?$

686

Threads

110K

Posts

910K

Credits

Credits
91229
QQ

Show all posts

kuing Posted 2024-4-9 14:01
hbghlyj 发表于 2024-4-9 04:40
Plot3D[(1/6) (-8 x - 2 y + 9), {x, 3/8, 7/8}, {y, 0, 1}]
应该用 ContourPlot3D[4 x + y + 3 z == 9/2, {x, 0, 1}, {y, 0, 1}, {z, 0, 1}]

Comment

太高深了!  Posted 2024-4-9 14:18
画个图而已,高啥深😓  Posted 2024-4-9 15:10

686

Threads

110K

Posts

910K

Credits

Credits
91229
QQ

Show all posts

kuing Posted 2024-4-9 15:19
jhsinutopia 发表于 2024-4-9 07:11
its typo, sorry! my mistake. I want
$\min((1-\alpha)\beta+(1-\beta)\gamma+(1-\gamma)\alpha)=?$
那就直接代 `\beta=9/2-4\alpha-3\gamma` 然后配方
\[(1-\alpha)\beta+(1-\beta)\gamma+(1-\gamma)\alpha=4\left(\alpha+\frac34\gamma-\frac{15}{16}\right)^2+\frac34\left(\gamma-\frac7{12}\right)^2+\frac{35}{48}\geqslant\frac{35}{48},\]
当 `\alpha=1/2`, `\beta=3/4`, `\gamma=7/12` 时取等。

3159

Threads

7941

Posts

610K

Credits

Credits
63770
QQ

Show all posts

hbghlyj Posted 2024-4-9 16:27
kuing 发表于 2024-4-9 07:19
当 `\alpha=1/2`, `\beta=3/4`, `\gamma=7/12` 时取等。
也可以用3#的方法,wolframalpha.com/input?i=Inverse[{{0,1,1,4},{ … .{{1},{1},{1},{9/2}}
$\pmatrix{y\\\lambda}=\pmatrix{C^{-1}&A\\A^\top&0}^{-1}\pmatrix{b\\f}=\left(\begin{array}{llll}
0 & 1 & 1 & 4 \\
1 & 0 & 1 & 1 \\
1 & 1 & 0 & 3 \\
4 & 1 & 3 & 0
\end{array}\right)^{-1} \cdot\left(\begin{array}{l}
1 \\
1 \\
1 \\
\frac{9}{2}
\end{array}\right)=\frac{1}{12}\left(\begin{array}{c}
6 \\
9 \\
7 \\
-1
\end{array}\right)$
即$\alpha=\frac6{12},\beta=\frac9{12},\gamma=\frac7{12}$

686

Threads

110K

Posts

910K

Credits

Credits
91229
QQ

Show all posts

kuing Posted 2024-4-9 17:22
也可以这样,令
\[(\alpha,\beta,\gamma)=\left(\frac x8+\frac12,\frac y2+\frac12,\frac z6+\frac12\right),\]

\[4\alpha+\beta+3\gamma=\frac92\iff x+y+z=1,\]

\[(1-\alpha)\beta+(1-\beta)\gamma+(1-\gamma)\alpha=\frac34-\frac{4yz+zx+3xy}{48},\]
然后类似于这帖 forum.php?mod=viewthread&tid=10200#pid50465(15#)的方法:
在 `\triangle ABC` 中恒有 `(x+y+z)^2\geqslant4(yz\sin^2A+zx\sin^2B+xy\sin^2C)`。
只需找到一个三角形使得 `\sin^2A:\sin^2B:\sin^2C=4:1:3` 即可,同样注意到:`4=1+3`,所以显然 `A=90\du`,由此得 `\sin^2B=1/4`, `\sin^2C=3/4`,代回去就是
\[(x+y+z)^2\geqslant4yz+zx+3xy,\]
所以
\[(1-\alpha)\beta+(1-\beta)\gamma+(1-\gamma)\alpha\geqslant\frac34-\frac1{48}=\frac{35}{48}.\]

Mobile version|Discuz Math Forum

2025-5-31 11:13 GMT+8

Powered by Discuz!

× Quick Reply To Top Edit