Forgot password?
 Create new account
View 435|Reply 27

[不等式] 柯西不等式的加强

[Copy link]

62

Threads

180

Posts

1289

Credits

Credits
1289

Show all posts

nttz Posted at 2022-9-12 11:49:09 |Read mode

Related collections:

801

Threads

4889

Posts

310K

Credits

Credits
36169

Show all posts

isee Posted at 2022-9-12 11:56:06 From the mobile phone
最右括号就是零,吓唬人的

Comment

??哪有零?  Posted at 2022-9-12 12:09

62

Threads

180

Posts

1289

Credits

Credits
1289

Show all posts

 Author| nttz Posted at 2022-9-12 17:07:51
isee 发表于 2022-9-12 11:56
最右括号就是零,吓唬人的
如何证明

Comment

我错了  Posted at 2022-9-12 20:26

701

Threads

110K

Posts

910K

Credits

Credits
94172
QQ

Show all posts

kuing Posted at 2022-9-12 17:13:04
想是不难想,难在后面怎么表达才顺畅……

由拉格朗恒等式可知原不等式等价于
\[\sum_{1\leqslant i<j\leqslant10}(a_ib_j-a_jb_i)^2\geqslant(a_1b_2-a_2b_1+a_3b_4-a_4b_3+\cdots+a_9b_{10}-a_{10}b_9)^2,\]
为方便码字引入记号,用 `f(i,j)` 表示 `a_ib_j-a_jb_i`,下面证明:对任意正整数 `n`,都有
\[\sum_{1\leqslant i<j\leqslant2n}f(i,j)^2\geqslant\bigl(f(1,2)+f(3,4)+\cdots+f(2n-1,2n)\bigr)^2.\]

注意到恒等式
\[(a-b)(c-d)=(a-c)(b-d)-(a-d)(b-c),\]
由此可得
\[f(i,j)f(p,q)=f(i,p)f(j,q)-f(i,q)f(j,p),\]
于是有
\[2f(i,j)f(p,q)\leqslant f(i,p)^2+f(j,q)^2+f(i,q)^2+f(j,p)^2,\]
因此
\begin{align*}
&\bigl(f(1,2)+f(3,4)+\cdots+f(2n-1,2n)\bigr)^2\\
={}&f(1,2)^2+f(3,4)^2+\cdots+f(2n-1,2n)^2+\sum_{1\leqslant i<p\leqslant n}2f(2i-1,2i)f(2p-1,2p)\\
\leqslant{}& f(1,2)^2+f(3,4)^2+\cdots+f(2n-1,2n)^2\\
&+\sum_{1\leqslant i<p\leqslant n}\bigl( f(2i-1,2p-1)^2+f(2i,2p)^2+f(2i-1,2p)^2+f(2i,2p-1)^2 \bigr),
\end{align*}
考察最后这一式,首先证明它不存在重复的项:

显然求和符号的那些不会重复,要有重复是与前面的重复,那也只能是 `f(2i-1,2p)^2` 与前面的重复,但是 `2p-(2i-1)=2(p-i)+1\geqslant3`,所以还是不会重复,所以没有重复(表达能力跟不上,见谅)。

然后,它的项数为 `n+4C_n^2`,正好等于 `C_{2n}^2`。

综合这两点,就说明了最后这一式实际上就是 `\sum_{1\leqslant i<j\leqslant2n}f(i,j)^2`,即得证。

701

Threads

110K

Posts

910K

Credits

Credits
94172
QQ

Show all posts

kuing Posted at 2022-9-12 17:54:18
将“于是有”那里的放缩改写为配方式,就可以得到原题的装X恒等式:
\begin{align*}
&(a_1^2+a_2^2+\cdots+a_{2n}^2)(b_1^2+b_2^2+\cdots+b_{2n}^2)\\
={}&(a_1b_1+a_2b_2+\cdots+a_{2n}b_{2n})^2\\
&+(a_1b_2-a_2b_1+a_3b_4-a_4b_3+\cdots+a_{2n-1}b_{2n}-a_{2n}b_{2n-1})^2\\
&+\sum_{1\leqslant i<p\leqslant n}(a_{2i-1}b_{2p-1}-a_{2p-1}b_{2i-1}-a_{2i}b_{2p}+a_{2p}b_{2i})^2\\
&+\sum_{1\leqslant i<p\leqslant n}(a_{2i-1}b_{2p}-a_{2p}b_{2i-1}+a_{2i}b_{2p-1}-a_{2p-1}b_{2i})^2.
\end{align*}

801

Threads

4889

Posts

310K

Credits

Credits
36169

Show all posts

isee Posted at 2022-9-12 19:42:20
isee 发表于 2022-9-12 11:56
最右括号就是零,吓唬人的
由 Cauchy 不等式$\sum a^2\sum b^2\geqslant (\sum ab)^2$,此时 $a_1b_2-a_2b_1=0,\cdots,a_9a_{10}-a_{10}b_9=0$,于是$\sum a^2\sum b^2\geqslant (\sum ab)^2+0$,逻辑错了?
isee=freeMaths@知乎

701

Threads

110K

Posts

910K

Credits

Credits
94172
QQ

Show all posts

kuing Posted at 2022-9-12 20:13:39
isee 发表于 2022-9-12 19:42
由 Cauchy 不等式$\sum a^2\sum b^2\geqslant (\sum ab)^2$,此时 $a_1b_2-a_2b_1=0,\cdots,a_9a_{10}-a_{10}b_9=0$,于是$\sum a^2\sum b^2\geqslant (\sum ab)^2+0$,逻辑错了?
没想到你也会有这种逻辑错😂
如果按这种逻辑,那最后那平方项的系数就可以任意大了

801

Threads

4889

Posts

310K

Credits

Credits
36169

Show all posts

isee Posted at 2022-9-12 20:25:45
kuing 发表于 2022-9-12 20:13
没想到你也会有这种逻辑错😂
如果按这种逻辑,那最后那平方项的系数就可以任意大了 ...
正常啦,看到题,10秒内的反应反映~~

Comment

是不是想成了取等条件?  Posted at 2022-9-12 23:24
isee=freeMaths@知乎

62

Threads

180

Posts

1289

Credits

Credits
1289

Show all posts

 Author| nttz Posted at 2022-9-12 22:14:30
isee 发表于 2022-9-12 20:25
正常啦,看到题,10秒内的反应反映~~
你用结果来证明条件?

62

Threads

180

Posts

1289

Credits

Credits
1289

Show all posts

 Author| nttz Posted at 2022-9-12 22:17:46
kuing 发表于 2022-9-12 17:13
想是不难想,难在后面怎么表达才顺畅……

由拉格朗恒等式可知原不等式等价于
这个也超级难看懂啊

62

Threads

180

Posts

1289

Credits

Credits
1289

Show all posts

 Author| nttz Posted at 2022-9-12 22:19:44
kuing 发表于 2022-9-12 17:13
想是不难想,难在后面怎么表达才顺畅……

由拉格朗恒等式可知原不等式等价于
能说明下拉格朗日恒等式是什么?如何证明么?

701

Threads

110K

Posts

910K

Credits

Credits
94172
QQ

Show all posts

kuing Posted at 2022-9-12 23:18:55
nttz 发表于 2022-9-12 22:19
能说明下拉格朗日恒等式是什么?如何证明么?
百度一下不就有了:https://baike.baidu.com/item/拉格朗日恒等式/7350857

276

Threads

691

Posts

6120

Credits

Credits
6120

Show all posts

力工 Posted at 2022-9-12 23:27:14
kuing 发表于 2022-9-12 23:18
百度一下不就有了:https://baike.baidu.com/item/拉格朗日恒等式/7350857
感觉有点行列式变形的味道,可是$(a_1^2+…+a_n^2)(b_1^2+…+b_n^2)$等等无法用行列式表示出来l

62

Threads

180

Posts

1289

Credits

Credits
1289

Show all posts

 Author| nttz Posted at 2022-9-13 09:19:29
kuing 发表于 2022-9-12 17:13
想是不难想,难在后面怎么表达才顺畅……

由拉格朗恒等式可知原不等式等价于
晕啊,每一步能详细点么,定义f(i,j),根据(a-b)(c-d)的恒等式如何得到下一步的,还有再下一步,理由是什么啊

701

Threads

110K

Posts

910K

Credits

Credits
94172
QQ

Show all posts

kuing Posted at 2022-9-13 14:00:55
nttz 发表于 2022-9-13 09:19
晕啊,每一步能详细点么,定义f(i,j),根据(a-b)(c-d)的恒等式如何得到下一步的,还有再下一步, ...
对 `(a-b)(c-d)=(a-c)(b-d)-(a-d)(b-c)` 作置换 `(a,b,c,d)\to(a_i/b_i,a_j/b_j,a_p/b_p,a_q/b_q)` 即得。

再下一步的理由就是 `x^2+y^2\geqslant\pm2xy` 啊

62

Threads

180

Posts

1289

Credits

Credits
1289

Show all posts

 Author| nttz Posted at 2022-9-13 18:41:51
kuing 发表于 2022-9-13 14:00
对 `(a-b)(c-d)=(a-c)(b-d)-(a-d)(b-c)` 作置换 `(a,b,c,d)\to(a_i/b_i,a_j/b_j,a_p/b_p,a_q/b_q)` 即得 ...
这个置换如何想到的?太神奇了吧

3146

Threads

8493

Posts

610K

Credits

Credits
66158
QQ

Show all posts

hbghlyj Posted at 2022-9-14 08:26:22
力工 发表于 2022-9-12 16:27
感觉有点行列式变形的味道,可是$(a_1^2+…+a_n^2)(b_1^2+…+b_n^2)$等等无法用行列式表示出来l ...
补上链接
kuing.cjhb.site/forum.php?mod=viewthread&tid=416

3146

Threads

8493

Posts

610K

Credits

Credits
66158
QQ

Show all posts

hbghlyj Posted at 2022-9-14 08:39:41
This is Cauchy's inequality for complex vector space.
$$(|z_1|^2+\dots+|z_5|^2)(|w_1|^2+\dots+|w_5|^2)\ge|z_1\bar{w_1}+\dots+z_5\bar{w_5}|^2$$
where $z_1=a_1+ia_2,\dots,z_5=a_9+ia_{10},w_1=b_1+ib_2,\dots,b_5=b_9+b_{10}$.

3146

Threads

8493

Posts

610K

Credits

Credits
66158
QQ

Show all posts

hbghlyj Posted at 2022-9-14 08:41:37
brilliant.org/wiki/cauchy-schwarz-inequality/ … rm-of-cauchy-schwarz
For a complex vector space, we have\[ \left| \sum_{i=1}^n x_i \bar{y}_i \right|^2 \leq \sum_{j=1}^n |x_j|^2 \sum_{k=1}^n |y_k|^2.\]Equality holds if and only if $x$ and $y$ are linearly dependent, that is, one is a scalar multiple of the other (which includes the case when one or both are zero).

3146

Threads

8493

Posts

610K

Credits

Credits
66158
QQ

Show all posts

hbghlyj Posted at 2022-9-14 08:53:28
All Lectures.pdf page77 (no text layer)(Producer: Microsoft Print to PDF)(File size: 68 MB)(Printing PDF causes file size to explode)
Define the inner product on $\Bbb C^n$ to be$$\left(z_{1}, \ldots, z_{n}\right) \cdot\left(w_{1}, \ldots, w_{n}\right)=\sum_{i=1}^{n} z_{i} \overline{w_{i}}$$We then have that$$\left(z_{1}, \ldots, z_{n}\right) \cdot\left(z_{1}, \ldots, z_{n}\right)=\sum_{i=1}^{n} z_{i} \overline{z_{i}}=\sum_{i=1}^{n}\left|z_{i}\right|^{2}$$which is positive-definite.

Hermitian Inner Product - mathworld
The basic example is the form\[h(z, w)=\sum z_{i} \overline{w_i}\]on $\Bbb C^n$, where $z=(z_1,...z_n)$ and $w=(w_1,...,w_n)$. Note that by writing $z_k=x_k+iy_k$, it is possible to consider $\Bbb C^n\sim\Bbb R^{2n}$, in which case $\Re[h]$ is the Euclidean inner product and $\Im[h]$ is a nondegenerate alternating bilinear form, i.e., a symplectic form. Explicitly, in $\Bbb C^2$, the standard Hermitian form is expressed below.\begin{aligned}h\left(\left(z_{11}, z_{12}\right),\left(z_{21}, z_{22}\right)\right)=&x_{11} x_{21}+x_{12} x_{22}+y_{11} y_{21}+y_{12} y_{22}\\&+i\left(x_{21} y_{11}-x_{11} y_{21}+x_{22} y_{12}-x_{12} y_{22}\right)\end{aligned}

手机版Mobile version|Leisure Math Forum

2025-4-20 22:09 GMT+8

Powered by Discuz!

× Quick Reply To Top Return to the list