<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="3.10.0">Jekyll</generator><link href="https://baoyu.space/feed.xml" rel="self" type="application/atom+xml" /><link href="https://baoyu.space/" rel="alternate" type="text/html" /><updated>2025-12-15T04:46:01+00:00</updated><id>https://baoyu.space/feed.xml</id><title type="html">Bao Yu</title><subtitle>Stay Curious.</subtitle><author><name>Bao Yu</name><email>baoyudu@outlook.com</email></author><entry><title type="html">Understanding the Concept of Conditional Expectation</title><link href="https://baoyu.space/Conditional-Expectation-1/" rel="alternate" type="text/html" title="Understanding the Concept of Conditional Expectation" /><published>2023-06-08T00:00:00+00:00</published><updated>2023-06-08T00:00:00+00:00</updated><id>https://baoyu.space/Conditional-Expectation-1</id><content type="html" xml:base="https://baoyu.space/Conditional-Expectation-1/"><![CDATA[<p>A elementary conditional expectations such as $E[X \mid Y=2]$  are numbers. If we consider $E[X \mid Y=y]$, it is a number that depends on $y$. So it is a function of $y$. Thus we can condition with respect to a σ-algebra, and view the conditional expectation itself as arandom variable.</p>
<h2 id="conditional-probability-function">Conditional Probability Function</h2>
<p>Let’s consider two <strong>discrete</strong> random variables $X$ and $Y$.<br />
Let $p(x, y)=\mathrm{P}(X=x, Y=y)$ be the joint probability mass function, then the marginal distribution</p>

\[p_X(x)=\mathrm{P}(X=x)=\sum_{y \in B} p(x, y),\]

<p>where $B$ is the set of possible values of $Y$.<br />
Similarly,</p>

\[p_Y(y)=\mathrm{P}(Y=y)=\sum_{x \in A} p(x, y),\]

<p>where $A$ is the set of possible values of $X$.<br />
Then the conditional probability mass function of $X$ given $Y=y$ is</p>

\[p_{X \mid Y}(x \mid y)=\mathrm{P}(X=x \mid Y=y)=\frac{p(x, y)}{p_Y(y)} .\]

<h2 id="conditional-expectation">Conditional Expectation</h2>
<h3 id="elementary-version">Elementary Version</h3>
<p>The conditional expectation of $X$ given $Y=y$ is defined as</p>

\[\mathrm{E}[X \mid Y=y]=\sum_{x \in A} x p_{X \mid Y}(x \mid y) .\]

<p>Consider a real-valued function $h$ from $\mathcal{R}$ to $\mathcal{R}$.<br />
The conditional expectation of $h(X)$ given $Y=y$ is</p>

\[\mathrm{E}[h(X) \mid Y=y]=\sum_{x \in A} h(x) p_{X \mid Y}(x \mid y) .\]

<p>The conditional expectation of $X$ given $Y$, denoted by $\mathrm{E}[X \mid Y]$, is the function of $Y$ that is defined to be $\mathrm{E}[X \mid Y=y]$ when $Y=y$.<br />
Specifically, let $\delta(x)$ be the function with $\delta(0)=1$ and $\delta(x)=0$ for all $x \neq 0$.<br />
Also, let $\delta_y(Y)=\delta(Y-y)$ be the indicator random variable such that $\delta_y(Y)=1$ if the event ${Y=y}$ occurs and $\delta_y(Y)=0$ otherwise.<br />
Then</p>

<p>\(\mathrm{E}[X \mid Y]=\sum_{y \in B} \mathrm{E}[X \mid Y=y] \delta_y(Y)=\sum_{y \in B} \sum_{x \in A} x p_{X \mid Y}(x \mid y) \delta_y(Y) .\)</p>
<blockquote>
  <p>That is to say:  The conditional expectation of $X$ give $Y$ is a random variable that takes in an event, and output a conditional expectation of $X$ given the value of $Y = y$ on that event</p>
</blockquote>

<h3 id="general-version">General Version</h3>

<blockquote class="def">
  <p>For a $\sigma$-algebra $\mathcal{G}, \mathrm{E}[X \mid \mathcal{G}]$ is defined to be the random variable that satisfies</p>
  <ol>
    <li>$\mathrm{E}[X \mid \mathcal{G}]$ is $\mathcal{G}$-measurable</li>
    <li>$\int_A X d \mathrm{P}=\int_A \mathrm{E}[X \mid \mathcal{G}] d \mathrm{P}$ for all $A \in \mathcal{G}$.</li>
  </ol>
</blockquote>

<p>To understand this definition, consider the $\sigma$-algebra generated by the random variable $Y$ (denoted by $\sigma(Y)$ ).</p>

<p>The condition that $\mathrm{E}[X \mid Y]$ is $\sigma(Y)$-measurable is simply that $\mathrm{E}[X \mid Y]$ is a measurable function of $Y$, i.e., $\mathrm{E}[X \mid Y]=h(Y)$ for some measurable function.</p>

<p>To understand the second condition, we may rewrite it as follows:</p>

<p>\(\mathrm{E}\left[\mathbf{1}_A X\right]=\mathrm{E}\left[\mathbf{1}_A \mathrm{E}[X \mid Y]\right],\)<br />
for all event $A$ in $\sigma(Y)$, where $\mathbf{1}_A$ is the indicator random variable with $\mathbf{1}_A=1$ when the event $A$ occurs.</p>]]></content><author><name>Bao Yu</name><email>baoyudu@outlook.com</email></author><summary type="html"><![CDATA[A elementary conditional expectations such as $E[X \mid Y=2]$ are numbers. If we consider $E[X \mid Y=y]$, it is a number that depends on $y$. So it is a function of $y$. Thus we can condition with respect to a σ-algebra, and view the conditional expectation itself as arandom variable.]]></summary></entry><entry><title type="html">Understand the concept of Filtration</title><link href="https://baoyu.space/Filtration/" rel="alternate" type="text/html" title="Understand the concept of Filtration" /><published>2023-06-05T16:41:00+00:00</published><updated>2023-06-05T16:41:00+00:00</updated><id>https://baoyu.space/Filtration</id><content type="html" xml:base="https://baoyu.space/Filtration/"><![CDATA[<blockquote class="def">
  <p>A filtration on $(\Omega, \mathcal{A}, \mathbb{P})$ is an increasing sequence $\left(\mathcal{F}<em>n\right)</em>{n \in \mathbb{Z}_{+}}$of sub- $\sigma$-fields of $\mathcal{A}$. We have thus<br />
\(\mathcal{F}_0 \subset \mathcal{F}_1 \subset \mathcal{F}_2 \subset \cdots \subset \mathcal{A}\)</p>
</blockquote>

<p>We also say that $\left(\Omega, \mathcal{A},\left(\mathcal{F}<em>n\right)</em>{n \in \mathbb{Z}<em>{+}}, \mathbb{P}\right)$ is a filtered probability space.<br />
The parameter $n \in \mathbb{Z}</em>{+}$is usually interpreted as time. The $\sigma$-field $\mathcal{F}_n$ gathers the information available at time $n$ (events that are $\mathcal{F}_n$-measurable are interpreted as those that depend only on what has happened up to time $n$ ). We will write<br />
\(\mathcal{F}_{\infty}=\bigvee_{n \in \mathbb{Z}_{+}} \mathcal{F}_n:=\sigma\left(\bigcup_{n \in \mathbb{Z}_{+}} \mathcal{F}_n\right)\)</p>
<h2 id="examples">Examples</h2>
<p>(a) If $\left(X_n\right)<em>{n \in \mathbb{Z}</em>{+}}$is a random process, then, for every $n \in \mathbb{Z}<em>{+}$, we may define $\mathcal{F}_n^X$ as the smallest $\sigma$-field on $\Omega$ for which $X_0, X_1, \ldots, X_n$ are measurable:<br />
\(\mathcal{F}_n^X=\sigma\left(X_0, X_1, \ldots, X_n\right) .\)<br />
Then $\left(\mathcal{F}_n^X\right)</em>{n \in \mathbb{Z}<em>{+}}$is a filtration called the canonical filtration of $\left(X_n\right)</em>{n \in \mathbb{Z}<em>{+}}$.<br />
(b) Suppose that $\Omega=[0,1), \mathcal{A}$ is the Borel $\sigma$-field on $[0,1)$, and $\mathbb{P}$ is Lebesgue measure. For every $n \in \mathbb{Z}</em>{+}$, set<br />
\(\mathcal{F}_n=\sigma\left(\left[\frac{i-1}{2^n}, \frac{i}{2^n}\right) ; i \in\left\{1,2, \ldots, 2^n\right\}\right) .\)<br />
Then $\left(\mathcal{F}<em>n\right)</em>{n \in \mathbb{Z}_{+}}$is a filtration called the dyadic filtration on $[0,1)$.</p>

<h2 id="understanding">Understanding</h2>
<p>The $\sigma$-algebra $\mathcal{F}_t$ represents all the events that we can “know” or “measure” at time $t$. Here’s a way to think about this.</p>

<p>Consider a simple example. Suppose we are observing the weather each day. We might have a stochastic process ${X_t}$ where $X_t$ represents the weather (say, temperature) at day $t$.</p>

<p>Now, at any given day $t$, what do we know? Well, we certainly know the weather from all previous days - those are past events that we’ve observed. And we know the weather today. But we do not know the weather on future days.</p>

<p>So, the $\sigma$-algebra $\mathcal{F}_t$ includes all the events that involve the weather up to and including day $t$. If $E$ is an event in $\mathcal{F}_t$, that means $E$ is a set of outcomes that we can determine whether or not has occurred based on the weather up to day $t$. For example, the event “It was hotter than 80 degrees on day 3” is in $\mathcal{F}_t$ for any $t \geq 3$.</p>

<p>That’s why $\mathcal{F}_t$ is interpreted as “the information available at time $t$”. <strong>It includes all the events that we can know whether they have happened or not based on what we’ve observed up to time $t$.</strong></p>

<p>To help understand this concept, let’s look at a specific example:</p>

<p>Suppose we have a fair coin, where $X_n$ represents the result of the $n^{th}$ toss, and $F_1$ and $F_2$ are the sigma-algebras corresponding to the information we obtain after the first and second tosses, respectively.</p>

<p>We label the outcomes as “H” for heads (or 1) and “T” for tails (0).</p>

<p>Then, $\mathcal{F}_1$ would contain all events that can be determined from the first toss. Specifically, they can be enumerated as:</p>

<ul>
  <li>${\Phi}$: The empty set, corresponding to an event with no outcome (an impossible event).</li>
  <li>${H}$: The event where the first toss results in heads.</li>
  <li>${T}$: The event where the first toss results in tails.</li>
  <li>${H, T}$: The entire sample space, corresponding to a certain event.</li>
</ul>

<p>Similarly, $\mathcal{F}_2$ would contain all events that can be determined from the first two tosses. It can be enumerated as:</p>

<ul>
  <li>${\Phi}$: The empty set.</li>
  <li>${HH}$: The event where two consecutive tosses are heads.</li>
  <li>${HT}$: The event where the first toss is heads and the second is tails.</li>
  <li>${TH}$: The event where the first toss is tails and the second is heads.</li>
  <li>${TT}$: The event where two consecutive tosses are tails.</li>
  <li>${HH, HT}$: The event where the first toss results in heads (regardless of the second toss’s outcome).</li>
  <li>${TH, TT}$: The event where the first toss results in tails (regardless of the second toss’s outcome).</li>
  <li>${HH, TH}$: The event where the second toss results in heads (regardless of the first toss’s outcome).</li>
  <li>${HT, TT}$: The event where the second toss results in tails (regardless of the first toss’s outcome).</li>
  <li>${HH, HT, TH, TT}$: The entire sample space.</li>
</ul>

<p>The role of a $\sigma$-algebra in this context is to formalize the idea of “events” or “outcomes” that we can measure or know. It is a mathematical tool that lets us handle the concept of information in a precise way.</p>

<p>This understanding also leads to the idea of a process being adapted to a filtration.</p>

<blockquote class="def">
  <p>A stochastic process ${X_t}$ is adapted to a filtration ${\mathcal{F}_t}$ if the value $X_t$ at each time $t$ is a function of the information available at that time, i.e., is $\mathcal{F}_t$-measurable.</p>
</blockquote>

<p>This means that, for each $t$, the value $X_t$ can be determined based on the information available up to and including time $t$.</p>]]></content><author><name>Bao Yu</name><email>baoyudu@outlook.com</email></author><summary type="html"><![CDATA[A filtration on $(\Omega, \mathcal{A}, \mathbb{P})$ is an increasing sequence $\left(\mathcal{F}n\right){n \in \mathbb{Z}_{+}}$of sub- $\sigma$-fields of $\mathcal{A}$. We have thus]]></summary></entry><entry><title type="html">深入理解Box-Muller变换</title><link href="https://baoyu.space/Box-Muller-Transform/" rel="alternate" type="text/html" title="深入理解Box-Muller变换" /><published>2023-03-03T00:00:00+00:00</published><updated>2023-03-03T00:00:00+00:00</updated><id>https://baoyu.space/Box-Muller-Transform</id><content type="html" xml:base="https://baoyu.space/Box-Muller-Transform/"><![CDATA[<p>通过计算机程序模拟生成服从均匀分布非常容易，但我们更常使用的是正态分布随机数。Box-Muller变换可以高效地利用均匀分布随机数生成正态分布随机数。</p>

<p>假设$(X,Y)$是平面上的一个随机的点，并假设其坐标$X$和$Y$是独立的标准正态随机变量。我们感兴趣的是$R,\Theta$的联合分布，即$(x,y)$的极坐标表示。</p>

<p>先假设$X$和$Y$都是正数。对于正的$x$和$y$，令 $r=\sqrt{x^2+y^2}$ 和 $\theta=\tan^{-1}y/x$，我们可以看出</p>

\[\begin{aligned} \frac{\partial g_1}{\partial x} &amp; =\frac{x}{\sqrt{x^2+y^2}} \\ \frac{\partial g_1}{\partial y} &amp; =\frac{y}{\sqrt{x^2+y^2}} \end{aligned}\]

\[\begin{aligned}
 &amp; \frac{\partial g_2}{\partial x}=\frac{1}{1+(y / x)^2}\left(\frac{-y}{x^2}\right)=\frac{-y}{x^2+y^2} \\ &amp; \frac{\partial g_2}{\partial y}=\frac{1}{x\left[1+(y / x)^2\right]}=\frac{x}{x^2+y^2} 
 \end{aligned}\]

<p>因此，</p>

\[J(x, y)=\frac{x^2}{\left(x^2+y^2\right)^{3 / 2}}+\frac{y^2}{\left(x^2+y^2\right)^{3 / 2}}=\frac{1}{\sqrt{x^2+y^2}}=\frac{1}{r}\]

<p>由于在$X$和$Y$都为正的条件下，$X,Y$的条件联合密度函数是</p>

\[f(x, y \mid X&gt;0, Y&gt;0)=\frac{f(x, y)}{P(X&gt;0, Y&gt;0)}=\frac{2}{\pi} e^{-\left(x^2+y^2\right) / 2}, x&gt;0, y&gt;0\]

<p>我们可以看出当$X$和$Y$都为正时$R=\sqrt{X^2+Y^2}$ 和 $\Theta=\tan ^{-1}(Y / X)$的条件联合密度函数是</p>

\[f(r, \theta \mid X&gt;0, Y&gt;0)=\frac{2}{\pi} r e^{-r^{2 / 2}}, \quad 0&lt;\theta&lt;\pi / 2, \quad 0&lt;r&lt;\infty\]

<p>由于我们研究的情况具有对称性, 很容易验证上面的联合密度函数对$X$和$Y$不都为正时也成立, 于是有</p>

<p>\(f(r, \theta)=\frac{1}{2 \pi} r e^{-r^2 / 2} \quad 0&lt;\theta&lt;2 \pi, \quad 0&lt;r&lt;\infty\)<br />
 现在，这个联合密度函数可以分解为$R$和$\Theta$的边缘密度函数，因此$R$和$\Theta$是相互独立的随机变量，其中$\Theta$均匀地分布于 $(0,2 \pi)$，$R$具有参数为$\frac{1}{2}$的Rayleigh分布，其密度函数为</p>

\[f(r)=r e^{-r^2 / 2} \quad 0&lt;r&lt;\infty\]

<blockquote>
  <p>当一个弓箭手瞄准靶心时，如果水平和垂直的误差距离是独立的标准正态分布，误差的绝对值就服从Rayleigh分布。</p>
</blockquote>

<p>这是个很符合直觉的结论，若二维平面上点的横纵坐标服从$iid$的正态分布，其方位角不仅是均匀分布的，并且与向量到原点的距离无关。<br />
现在, 我们还想知道$R^2$和$\Theta$的联合分布，由于变换$d=g_1(x, y)=x^2+y^2$和$\theta=g_2(x, y)=\tan^{-1}y / x$的雅可比行列式为</p>

\[J=\left|
\begin{array}{cc} 2 x &amp; 2 y \\
\frac{-y}{x^2+y^2} &amp; \frac{x}{x^2+y^2} \end{array}
\right|=2\]

<p>因此，经过变量替换，我们得到 <br />
\(f(d, \theta)=\frac{1}{2} e^{-d / 2} \frac{1}{2 \pi} \quad 0&lt;d&lt;\infty, \quad 0&lt;\theta&lt;2 \pi\) 因此，$R^2$和$\Theta$是独立的，其中$R^2$具有参数为$\frac{1}{2}$的指数分布。</p>

<blockquote>
  <p>也可以利用卡方分布的构造, 根据$R^2=X^2+Y^2$，也可以发现$R^2$是具有2个自由度的卡方分布, 即指数分布。</p>
</blockquote>

<p>不要忘了我们的目标是把均匀分布转换为正态分布, 或者说, 用均匀分布把正态分布表示出来.<br />
既然满足正态分布的$X_1=R\cos\Theta$, 且$\Theta$服从$(0,2\pi)$上的均匀分布, 只剩服从指数分布的$R^2$需要处理, 也就是说, <strong>现在我们只需要把均匀分布转换为指数分布</strong>.</p>

<p>设$U_1$和$U_2$是独立随机变量，服从区间$(0,1)$上的均匀分布。很容易验证对于$x&gt;0$，$-2 \log U_1$服从指数分布，因为</p>

\[\begin{aligned} 
P\left(-2 \log U_1&lt;x\right) &amp; =P\left(\log U_1&gt;-\frac{x}{2}\right) \\ &amp; =P\left(U_1&gt;e^{-x / 2}\right) \\ &amp; =1-e^{-x / 2} 
\end{aligned}\]

<p>而$2\pi U_2$是一个$(0,2\pi)$上的均匀分布的随机变量，我们可以用它来生成$\Theta$。令</p>

\[\begin{aligned} &amp; R^2=-2 \log U_1 \\ &amp; \Theta=2 \pi U_2 \end{aligned}\]

<p>现在，由于$X_1=R\cos\Theta, X_2=R\sin\Theta$，可以得到</p>

\[\begin{aligned} 
&amp; X_1=\sqrt{-2 \log U_1} \cos \left(2 \pi U_2\right) \\ &amp; X_2=\sqrt{-2 \log U_1} \sin \left(2 \pi U_2\right) 
\end{aligned}\]

<p>是独立的标准正态分布随机变量。</p>]]></content><author><name>Bao Yu</name><email>baoyudu@outlook.com</email></author><summary type="html"><![CDATA[通过计算机程序模拟生成服从均匀分布非常容易，但我们更常使用的是正态分布随机数。Box-Muller变换可以高效地利用均匀分布随机数生成正态分布随机数。]]></summary></entry><entry><title type="html">Markdown渲染器与MathJax数学公式显示的严重冲突及解决方案</title><link href="https://baoyu.space/Jekyll-Math/" rel="alternate" type="text/html" title="Markdown渲染器与MathJax数学公式显示的严重冲突及解决方案" /><published>2023-01-04T00:00:00+00:00</published><updated>2023-01-04T00:00:00+00:00</updated><id>https://baoyu.space/Jekyll-Math</id><content type="html" xml:base="https://baoyu.space/Jekyll-Math/"><![CDATA[<p>令人惊讶的是，在Jekyll或其他任何静态博客中，Markdown这一排版工具与LaTex数学公式从来都是不合适的一对儿。<br />
在静态页面生成时，Markdown渲染器按照语法将原始的<code class="language-plaintext highlighter-rouge">.md</code>文件转换为HTML形式，在这一过程中，每两个下划线”_“会被转换成一对<code class="language-plaintext highlighter-rouge">&lt;em&gt;</code> <code class="language-plaintext highlighter-rouge">&lt;/em&gt;</code>标签, 导致页面加载时Mathjax无法正确识别LaTex公式中表示下标的下划线。这是一个广泛存在、并且在Jekyll社区、Commonmark 社区、StackOverFlow、以及Kramdown、Jekyll、MathJax等项目的Github Issue中均得到大量讨论的问题，MathJax的开发者Davide Cervone也参与了问题的讨论, 但从未有人给出过真正的解决方案。</p>

<p>MathJax文档中有这样一段：</p>

<blockquote>
  <p>Such systems need to be told not to modify the mathematics that appears between math delimiters. That usually involves modifying the content-management system itself, which is beyond the means of most page authors. If you are lucky, someone else will already have done this for you, and you may be able to find a MathJax plugin for your system using a web search.<br />
If there is no plugin for your system, or if the plugin doesn’t handle the subtleties of isolating the mathematics from the other markup that it supports, then you may have to “trick” the content-management system into leaving your mathematics untouched. Most content-management systems provide some means of indicating text that should not be modified (“verbatim” text), often for giving code snippets for computer languages. You may be able use that to enclose your mathematics so that the system leaves it unchanged and MathJax can process it.</p>
</blockquote>

<p>由于可用的Markdown渲染器(Kramdown、redCarpet、rDiscount等)均没有针对此问题的插件或设置项，网络上给出的所有临时解决方案也基本采用了这种思路, 均为用代码块或者<code class="language-plaintext highlighter-rouge">&lt;div&gt;</code>标签, 或者Liquid中的<code class="language-plaintext highlighter-rouge">{% raw %}</code> 标签等环境手动包裹住所有的数学表达式, 从而让Markdown渲染器跳过数学表达式部分,避免其破坏原有公式。但这是一个相当敷衍的办法，经常使用数学公式的作者需要不断地做额外的工作来正确显示公式，并且严重地影响了文章的可迁移性。</p>

<h2 id="我的解决方案">我的解决方案</h2>

<p>既然没有现成的插件或配置方法,我只能自己想想办法。基于一种新思路，我基本上完美解决了这个问题，公式可以正确显示，且不需要对原文档做额外修改。<br />
其原理为，在页面的<code class="language-plaintext highlighter-rouge">footer</code>部分插入一段JavaScript片段, 把所有<code class="language-plaintext highlighter-rouge">&lt;em&gt;</code> 和<code class="language-plaintext highlighter-rouge">&lt;/em&gt;</code>标签重新转换回”_”.</p>

<p>这里以Jekyll为例, 也可以类似应用在其他静态页面生成器中.<br />
在 _includes 文件夹中新建一个math.html文件,内容为对MathJax的加载:</p>

<div class="language-plaintext highlighter-rouge"><div class="highlight"><pre class="highlight"><code>&lt;script&gt;
    MathJax = {
      loader: {load: ['[tex]/ams']},
      tex: {
        autoload: {
          action: ['toggle', 'mathtip', 'texttip'],
          amscd: [[], ['CD']],
          bbox: ['bbox'],
          boldsymbol: ['boldsymbol'],
          braket: ['bra', 'ket', 'braket', 'set', 'Bra', 'Ket', 'Braket', 'Set', 'ketbra', 'Ketbra'],
          cancel: ['cancel', 'bcancel', 'xcancel', 'cancelto'],
          color: ['color', 'definecolor', 'textcolor', 'colorbox', 'fcolorbox'],
          enclose: ['enclose'],
          extpfeil: ['xtwoheadrightarrow', 'xtwoheadleftarrow', 'xmapsto',
                    'xlongequal', 'xtofrom', 'Newextarrow'],
          html: ['href', 'class', 'style', 'cssId'],
          mhchem: ['ce', 'pu'],
          newcommand: ['newcommand', 'renewcommand', 'newenvironment', 'renewenvironment', 'def', 'let'],
          unicode: ['unicode'],
          verb: ['verb']
        },
        inlineMath: [['$', '$'],['\\(','\\)']],
        displayMath: [['$$', '$$'],['\\[','\\]']],
        processEscapes: true,
        packages: {'[+]': ['ams']}
        }
    };

  &lt;/script&gt;

&lt;script type="text/javascript" id="MathJax-script" async
  src="https://cdn.jsdelivr.net/npm/mathjax@3.0.0/es5/tex-chtml.js"&gt;
&lt;/script&gt;
</code></pre></div></div>

<p>再新建一个replace.html ,插入JavaScript来实现内容的替换:</p>

<div class="language-javascript highlighter-rouge"><div class="highlight"><pre class="highlight"><code><span class="o">&lt;</span><span class="nx">script</span><span class="o">&gt;</span>
    <span class="kd">function</span> <span class="nx">rep</span><span class="p">()</span> 
    <span class="p">{</span>
        <span class="nb">document</span><span class="p">.</span><span class="nx">body</span><span class="p">.</span><span class="nx">innerHTML</span>
            <span class="o">=</span> <span class="nb">document</span><span class="p">.</span><span class="nx">body</span><span class="p">.</span><span class="nx">innerHTML</span>
            <span class="p">.</span><span class="nx">replaceAll</span><span class="p">(</span><span class="dl">"</span><span class="s2">&lt;em&gt;</span><span class="dl">"</span><span class="p">,</span> <span class="dl">"</span><span class="s2">_</span><span class="dl">"</span><span class="p">).</span><span class="nx">replaceAll</span><span class="p">(</span><span class="dl">"</span><span class="s2">&lt;/em&gt;</span><span class="dl">"</span><span class="p">,</span> <span class="dl">"</span><span class="s2">_</span><span class="dl">"</span><span class="p">)</span>
        
    <span class="p">}</span>
    <span class="nx">rep</span><span class="p">()</span>
<span class="o">&lt;</span><span class="sr">/script</span><span class="err">&gt;
</span></code></pre></div></div>

<p>由于Jekyll使用Liquid来渲染页面 ,我们在footer.html中利用Liquid Tag 来引用上述脚本, 注意math.html必须在replace.html之后引用.</p>

<div class="language-plaintext highlighter-rouge"><div class="highlight"><pre class="highlight"><code>
{% if page.math %}  
    {% include replace.html %}
    {% include math.html %} 
{% endif %}
</code></pre></div></div>

<p>这样, 在文章的Frontmatter中加上 <code class="language-plaintext highlighter-rouge">math: true</code>, 即可正确显示数学公式.</p>]]></content><author><name>Bao Yu</name><email>baoyudu@outlook.com</email></author><summary type="html"><![CDATA[令人惊讶的是，在Jekyll或其他任何静态博客中，Markdown这一排版工具与LaTex数学公式从来都是不合适的一对儿。]]></summary></entry><entry><title type="html">最容易证明的强大数定律</title><link href="https://baoyu.space/Fourth-Moment-SLLN/" rel="alternate" type="text/html" title="最容易证明的强大数定律" /><published>2022-12-01T00:00:00+00:00</published><updated>2022-12-01T00:00:00+00:00</updated><id>https://baoyu.space/Fourth-Moment-SLLN</id><content type="html" xml:base="https://baoyu.space/Fourth-Moment-SLLN/"><![CDATA[<p>四阶矩强大数定律虽然对“有限$4$阶矩”的条件作了限制，但并没有假设$\left(X_n\right)$序列的同分布, 而且这个很好的结论并不需要复杂的证明.</p>

<blockquote class="thm">
  <p><strong>Theorem</strong><br />
假设$X_1, X_2, \ldots$是独立的随机变量，并且存在$[0,\infty)$中的一个常数$K$，使得：</p>

\[\mathrm{E}\left(X_k\right)=0, \quad \mathrm{E}\left(X_k^4\right) \leq K, \quad \forall k .\]

  <p>令 $S_n=X_1+X_2+\cdots+X_n$. 则</p>

\[\mathbf{P}\left(n^{-1} S_n \rightarrow 0\right)=1,\]

  <p>即</p>

\[S_n / n \rightarrow 0, a.s.\]

</blockquote>

<p>证明： 我们有</p>

\[\begin{aligned} 
\mathrm{E}\left(S_n^4\right) &amp; =\mathrm{E}\left[\left(X_1+X_2+\cdots+X_n\right)^4\right] \\ &amp; =\mathrm{E}\left(\sum_k X_k^4+6 \sum \sum_{i&lt;j} X_i^2 X_j^2\right), 
\end{aligned}\]

<p>因为，对于不同的$i, j, k$和$l$，</p>

\[\mathrm{E}\left(X_i X_j^3\right)=\mathrm{E}\left(X_i X_j^2 X_k\right)=\mathrm{E}\left(X_i X_j X_k X_l\right)=0,\]

<p>利用独立性和$\mathrm{E}\left(X_i\right)=0$。</p>

<p>注意$\mathrm{E}\left(X_j^4\right)&lt;\infty$意味着$\mathrm{E}\left(X_j^3\right)&lt;\infty$，(因为“$\mathcal{L}^p$范数的单调性”)。因此，$X_i$和$X_j^3$都在$\mathcal{L}^{\mathbf{1}}$中。</p>

<p>我们知道</p>

\[\left[\mathrm{E}\left(X_i^2\right)\right]^2 \leq \mathrm{E}\left(X_i^4\right) \leq K, \quad \forall i .\]

<p>因此，再次利用独立性，对于$i \neq j$，</p>

\[\mathrm{E}\left(X_i^2 X_j^2\right)=\mathrm{E}\left(X_i^2\right) \mathrm{E}\left(X_j^2\right) \leq K\]

<p>因此，</p>

\[\mathrm{E}\left(S_n^4\right) \leq n K+3 n(n-1) K \leq 3 K n^2,\]

<p>和</p>

\[\mathrm{E} \sum\left(S_n / n\right)^4 \leq 3 K \sum n^{-2}&lt;\infty\]

<p>因此$\sum\left(S_n / n\right)^4&lt;\infty, a.s.$，，且</p>

\[S_n / n \rightarrow 0, \quad a.s.\]

<blockquote class="thm">
  <p><strong>Corollary</strong><br />
如果定理中的条件$\mathrm{E}\left(X_k\right)=0$被替换为$\mathrm{E}\left(X_k\right)=\mu$，其中$\mu$是常数，则定理的结论为$n^{-1}S_n\rightarrow\mu , a.s.$ .</p>
</blockquote>

<p>证明： 显然，令$Y_k:=X_k-\mu$,就可以将将定理应用于序列$\left(Y_k\right)$，但我们需要</p>

\[\sup _k \mathrm{E}\left(Y_k^4\right)&lt;\infty\]

<p>这可以通过Minkowski不等式 \(\left\|X_k-\mu\right\|_4 \leq\left\|X_k\right\|_4+\|\mu\|\) 得到.</p>]]></content><author><name>Bao Yu</name><email>baoyudu@outlook.com</email></author><summary type="html"><![CDATA[四阶矩强大数定律虽然对“有限$4$阶矩”的条件作了限制，但并没有假设$\left(X_n\right)$序列的同分布, 而且这个很好的结论并不需要复杂的证明.]]></summary></entry><entry><title type="html">怎样在文章页面显示美观的Admonition</title><link href="https://baoyu.space/Admonition/" rel="alternate" type="text/html" title="怎样在文章页面显示美观的Admonition" /><published>2022-11-27T00:00:00+00:00</published><updated>2022-11-27T00:00:00+00:00</updated><id>https://baoyu.space/Admonition</id><content type="html" xml:base="https://baoyu.space/Admonition/"><![CDATA[<p>我经常使用Obsidian中的Admonition插件来美化笔记页面. 用这个插件可以重点显示文章中的定理、定义，对可读性和美观度的提升非常明显。</p>

<p><a href="https://indii.org/blog/admonitions-in-markdown/">Admonitions in Markdown, Working or failing gracefully across Apostrophe, Kramdown, and Jekyll. No plugins required.</a><br />
这篇文章中提出了一种在Kramdown渲染器下可以正常工作的解决方案. 经我的改进, 目前可以实现理想效果:</p>

<h3 id="示例">示例</h3>

<blockquote class="notice--success">
  <p>Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.</p>
</blockquote>

<blockquote class="notice--danger">
  <p>Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.</p>
</blockquote>

<blockquote class="notice--warning">
  <p>Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.</p>
</blockquote>

<blockquote class="notice--info">
  <p>Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.</p>
</blockquote>

<blockquote class="thm">
  <p>Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.</p>
</blockquote>

<blockquote class="def">
  <p>Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.</p>
</blockquote>

<blockquote class="def">

  <p>设 $f(z)=u(x, y)+\mathrm{i} v(x, y)$ 是定义在域 $D$ 上的函数, $z_{0}=x_{0}+\mathrm{i} y_{0} \in$ $D$. 我们说 $f$ 在 $z_{0}$ 处<strong>实可微</strong>, 是指 $u$ 和 $v$ 作为 $x, y$ 的二元函数在 $\left(x_{0}, y_{0}\right)$ 处可微.</p>
</blockquote>

<h3 id="实现方法">实现方法</h3>

<p>添加如下css代码</p>

<div class="language-plaintext highlighter-rouge"><div class="highlight"><pre class="highlight"><code>/* To add support of admonition */
blockquote {
  @extend .alert;
  @extend .alert-primary;
  @extend .pb-0;
  font-size:15px;

}

blockquote.success {
  @extend .alert-success;
  /* set a color for the blockquote */
  border-left-color: #5cb85c;
}
blockquote.success p:before {
  content: "✓ Success \a";
  font-weight:bold;
  font-size:16px;

}

blockquote.danger {
  @extend .alert-danger;
  /* set a color for the blockquote */
  border-left-color: #d9534f;
}
blockquote.danger p:before {
  content: "⊠ Danger ";
  font-weight:bold;
  font-size:16px;
}

blockquote.warning {
  @extend .alert-warning;
  /* set a color for the blockquote */
  border-left-color: #f0ad4e;
}
blockquote.warning p:before {
  content: "⚠ Warning ";
  font-weight:bold;
  font-size:16px;
}

blockquote.info {
  @extend .alert-info;
  /* set a color for the blockquote */
  border-left-color: #ffeb9d;
}
blockquote.info p:before {
  content: "ⓘ Further Information ";
  font-weight:bold;
  font-size:16px;
}

/* add a type of admonition: thm*/
blockquote.thm {
  @extend .alert-info;
  /* set a color for the blockquote:blue */
  border-left-color: #45cdff;
  /* change text color to black */
  color: #000;
  /* cancel italics */
  font-style: normal;
}

blockquote.thm p:before {
  content: "Theorem \a";
  font-weight:bold;
  font-size:16px;
  display: block;
}

/* add a type of admonition: def */
blockquote.def {
  @extend .alert-info;
  /* set a color for the blockquote:deep blue */
  border-left-color: #1a8cff;
  /* change text color to black */
  color: #000;
  /* cancel italics */
  font-style: normal;
}

blockquote.def p:before { 
  content: "Definition \a ";
  font-weight:bold;
  font-size:16px;
  display: block;
}

/*end of admonition*/
</code></pre></div></div>

<p>每个CSS代码块包含两个部分的样式：</p>

<p>1.扩展了CSS类“alert-success”等的样式，用“success”类来替代它，指定块引用元素的左边框颜色</p>

<p>2.在blockquote元素的p元素之前，添加一个带有如“✓ Success”文字内容的伪元素</p>

<p>为了在伪元素和p元素之间添加一个换行，尝试在伪元素的content属性中使用 \a 表示换行符,但换行符没有生效. 查阅相关资料后尝试使用 display: block 或 display: table 形式将其强制转换为块级元素即可.</p>]]></content><author><name>Bao Yu</name><email>baoyudu@outlook.com</email></author><summary type="html"><![CDATA[我经常使用Obsidian中的Admonition插件来美化笔记页面. 用这个插件可以重点显示文章中的定理、定义，对可读性和美观度的提升非常明显。]]></summary></entry><entry><title type="html">孤立奇点</title><link href="https://baoyu.space/Isolated-Singularities/" rel="alternate" type="text/html" title="孤立奇点" /><published>2022-07-03T00:00:00+00:00</published><updated>2022-07-03T00:00:00+00:00</updated><id>https://baoyu.space/Isolated-Singularities</id><content type="html" xml:base="https://baoyu.space/Isolated-Singularities/"><![CDATA[<p>Laurent级数是研究全纯函数在孤立奇点附近性质的有力工具.</p>

<h2 id="定义">定义</h2>

<h3 id="通过附近的极限情况来定义">通过附近的极限情况来定义</h3>

<p>如果 $f$ 在无心圆盘 ${z: 0&lt;|z-z_{0}|&lt;R}$ 中解析, 就称 $z_{0}$ 是 $f$ 的<strong>孤立奇点</strong>.<br />
$f$ 在孤立奇点 $z_{0}$ 附近可能有三种情形:</p>

<p>(1) $\lim_{z \rightarrow z_{0}} f(z)=a, a$ 是一有限数, 这时称 $z_{0}$ 是 $f$ 的<strong>可去奇点</strong>;</p>

<p>(2) $\lim_{z \rightarrow z_{0}} f(z)=\infty$, 这时称 $z_{0}$ 是 $f$ 的<strong>极点</strong>;</p>

<p>(3) $\lim_{z \rightarrow z_{0}} f(z)$ 不存在, 这时称 $z_{0}$ 是 $f$ 的<strong>本性奇点</strong>.</p>

<h3 id="通过laurent展开来定义">通过Laurent展开来定义</h3>

<p>i. 如果当 $n=-1,-2,-3, \cdots$ 时, $\alpha_{n}=0$, 那么我们说 $z_{0}$ 是函数 $f(z)$ 的<strong>可去奇点</strong>, 或者说 $f(z)$ 在 $z_{0}$ 有可去奇点. 这是因为令 $f\left(z_{0}\right)=\alpha_{0}$, 就得到在整个圆盘 $\mid z-$ $z_{0} \mid&lt;R$ 内解析的函数 $f(z)$.</p>

<p>ii. 如果只有有限个(至少一个) 整数 $n&lt;0$, 使得 $\alpha_{n} \neq 0$, 那么我们说 $z_{0}$ 是函 数 $f(z)$ 的<strong>极点</strong>. 设对于正整数 $m, \alpha_{-m} \neq 0$; 而当 $n&lt;-m$ 时, $\alpha_{n}=0$. 那么我们就说 $z_{0}$ 是 $f(z)$ 的 <strong>$m$ 阶极点</strong>.</p>

<p>iii. 如果有无限个整数 $n&lt;0$, 使得 $\alpha_{n} \neq 0$, 那么我们说 $z_{0}$ 是 $f(z)$ 的<strong>本质奇点</strong>.</p>

<blockquote>

  <p>0 分别是 $\frac{\sin z}{z}, \frac{\sin z}{z^{2}}$ 及 $\mathrm{e}^{\frac{1}{z}}$ 的可去奇点、单极点及 本质奇点.</p>

</blockquote>

<h2 id="性质">性质</h2>

<p>(Weierstrass) 设 $z_{0}$ 是 $f$ 的本性奇点,那么对任意 $A \in \mathbb{C}<em>{\infty}$, 必存在趋于 $z</em>{0}$ 的点列 $\left \{ z_{n}\right\}$, 使得 $\lim <em>{n \rightarrow \infty} f\left(z</em>{n}\right)=A$.</p>

<p>证:<br />
先设 $A=\infty$. 因为 $z_{0}$ 是 $f$ 的本性奇点, 故 $f$ 在 $z_{0}$ 附近无界. 于是对任意自然 数 $n$, 总能找到 $z_{n}$, 使得 $\left|z_{n}-z_{0}\right|&lt;\frac{1}{n}$, 但 $\left|f\left(z_{n}\right)\right|&gt;n$, 这说明 $\lim <em>{n \rightarrow \infty} f\left(z</em>{n}\right)=\infty$.</p>

<p>再设 $A$ 是一个有限数. <br />
令 $\varphi(z)=\frac{1}{f(z)-A}$, 我们证明 $\varphi$ 在 $z_{0}$ 的邻域中无界. 不然的话, $z_{0}$ 是 $\varphi$ 的可去奇点, 适当选择 $\varphi\left(z_{0}\right)$ 的值, 可使 $\varphi$ 在 $z_{0}$ 处全纯. 如果 $\varphi\left(z_{0}\right) \neq 0$, 则因 $f(z)=\frac{1}{\varphi(z)}+A, f$ 也在 $z_{0}$ 处全纯, 这不可能. 故必有$\varphi\left(z_{0}\right)=0$, 这说明 $z_{0}$ 是 $f$ 的极点, 也不可能. 所以, $\varphi$ 在 $z_{0}$ 的邻域中无界. 于是, 对任意自然数 $n$, 存在 $z_{n}$, 使得 $\left|z_{n}-z_{0}\right|&lt;\frac{1}{n}$, 但 $\frac{1}{|f(z)-A|}&gt;n$, 即 $|f(z)-A|&lt;\frac{1}{n}$. 这就证明了 $\lim <em>{n \rightarrow \infty} f\left(z</em>{n}\right)=A$.</p>

<p>后来,Picard 又证明了比 Weierstrass 定理更进一步的结果：</p>

<p>(Picard) 全纯函数在本性奇点的邻域内<strong>无穷多次地取到每个有穷复值</strong>, 最多只有一个例外.</p>

<p>例如, 考虑函数 $f(z)=\mathrm{e}^{\frac{1}{z}}$, 它在 $z=0$ 附近是全纯的. 若让 $z$ 沿着 $x$ 轴分别从 0 的 左边和右边趋于 0 , 可得</p>

\[\begin{aligned}
&amp;\lim _{z=x \rightarrow 0^{-}} \mathrm{e}^{\frac{1}{z}}=\lim _{x \rightarrow 0^{-}} \mathrm{e}^{\frac{1}{x}}=0, \\
&amp;\lim _{z=x \rightarrow 0^{+}} \mathrm{e}^{\frac{1}{z}}=\lim _{x \rightarrow 0^{+}} \mathrm{e}^{\frac{1}{x}}=\infty
\end{aligned}\]

<p>这说明 $\lim <em>{z \rightarrow 0} \mathrm{e}^{\frac{1}{z}}$ 不存在, 所以 $z=0$ 是 $\mathrm{e}^{\frac{1}{z}}$ 的本性奇点. 对于任意复数 $a \neq 0$, 若取 $z</em>{n}=(\log a+2 n \pi \mathrm{i})^{-1}$, 则 $f\left(z_{n}\right)=\mathrm{e}^{\log a+2 n \pi \mathrm{i}}=a$. 由于 $z_{n} \rightarrow 0$, 这说明 $\mathrm{e}^{\frac{1}{z}}$ 在 $z=0$ 的邻域中可以无穷多次地取到非零值 $a$, 但 0 是它的唯一的例外值.</p>

<h2 id="无穷远为孤立奇点">无穷远为孤立奇点</h2>

<p>如果 $f$ 在无穷远点的去心邻域(不包括无穷远点)$\{z: 0 \leqslant R&lt;\vert z\vert &lt;\infty\}$ 中全纯, 就称 $\infty$ 是 $f$ 的孤立奇点.</p>

<p>在这种情形下, 作变换 $z=\frac{1}{\zeta}$, 记</p>

\[g(\zeta)=f\left(\frac{1}{\zeta}\right)\]

<p>则 $g$ 在 $0&lt;\vert \zeta \vert &lt;\frac{1}{R}$ 中全纯, 即 $\zeta=0$ 是 $g$ 的孤立奇点.</p>

<p>如果 $\zeta=0$ 是 $g$ 的可去奇点、 $m$ 阶极点或本性奇点, 那么我们相应地称 $z=\infty$ 是 $f$ 的可去奇点、 $m$ 阶极点或本性奇点.</p>

<p>因为 $g$ 在原点的邻域中有 Laurent 展开:</p>

\[g(\zeta)=\sum_{n=-\infty}^{\infty} b_{n} \zeta^{n}, 0&lt;\vert \zeta \vert &lt;\frac{1}{R}\]

<p>所以 $f$ 在 $R&lt;\vert z \vert &lt;\infty$ 中有下面的 Laurent 展开:</p>

\[f(z)=\sum_{n=-\infty}^{\infty} a_{n} z^{n}\]

<p>其中 $a_{n}=b_{-n}, n=0, \pm 1, \cdots$.</p>

<h3 id="无穷为可去奇点">无穷为可去奇点</h3>

<p>如果 $z=\infty$ 是 $f$ 的可去奇点, 即 $\zeta=0$ 是 $g$ 的可去奇点, 因而 $b_{n}=0$ $(n=-1,-2, \cdots)$, 即$g$的负次幂的系数都为零, 所以 $f$ 只剩下解析部分和常数, 其Laurent 展开为</p>

\[f(z)=\sum_{n=0}^{\infty} a_{-n} z^{-n}\]

<h3 id="无穷为极点或本性奇点">无穷为极点或本性奇点</h3>

<p>如果 $z=\infty$ 分别是 $f$ 的 $m$ 阶极点或本性奇点, 那么 $f$ 在 $ R &lt; \vert z \vert &lt; \infty$ 中分别有下面的 Laurent 展开式:</p>

\[f(z)=a_{m} z^{m}+\cdots+a_{1}+a_{0}+a_{-1} z^{-1}+\cdots\]

<p>或</p>

\[f(z)=\cdots+a_{m} z^{m}+\cdots+a_{1} z+a_{0}+a_{-1} z^{-1}+\cdots .\]

<blockquote>

  <p>对于无穷远为孤立奇点的情况, 反过来称 $\sum_{n=1}^{\infty} a_{n} z^{n}$ 为 $f$ 的主要部分, $\sum_{n=0}^{\infty} a_{-n} z^{-n}$ 为 $f$ 的解析部分.</p>
</blockquote>]]></content><author><name>Bao Yu</name><email>baoyudu@outlook.com</email></author><summary type="html"><![CDATA[Laurent级数是研究全纯函数在孤立奇点附近性质的有力工具.]]></summary></entry><entry><title type="html">不可测集(Vatali Set)的构造</title><link href="https://baoyu.space/Vatali-Set/" rel="alternate" type="text/html" title="不可测集(Vatali Set)的构造" /><published>2021-09-08T00:00:00+00:00</published><updated>2021-09-08T00:00:00+00:00</updated><id>https://baoyu.space/Vatali-Set</id><content type="html" xml:base="https://baoyu.space/Vatali-Set/"><![CDATA[<h2 id="理想测度">理想测度</h2>
<p>在我们心里, 在实数域中最”美好”的测度应该具有这几条性质:</p>

<ol>
  <li>
    <p>$\lambda :{\mathscr P}(R) \to R{_ + } \cup \left\{ \infty  \right\}$</p>
  </li>
  <li>
    <p>$\lambda ([a,b]) = b - a$</p>
  </li>
  <li>
    <p>$\lambda (A + x) = \lambda (A)$ (平移不变)</p>
  </li>
  <li>
    <p>$\lambda (\mathop  \bigsqcup \limits_{j \ge 1} {A_j}) = \sum\limits_{j \ge 1} {\lambda ({A_j})}$ ($\sigma$-可加)</p>
  </li>
</ol>

<p>在这样的测度下, 是否有不可测的集合存在?</p>
<h2 id="一些定义">一些定义</h2>

<p>若$x,y \in R$，且$y - x \in Q$(有理数集)，则称$x \sim y$(x和y等价).记为$[x] = \left\{ {y \in R\;|\;y - x \in Q} \right\}$.</p>

<p>$\Lambda$：所有$R$中元素的<strong>等价类组成的集合</strong><br />
$\Omega$：只取$\Lambda$中每个等价类的一个元素组成的集合。且$\Omega \in (0,1)$。</p>

<p>用类似数论中的语言, $\Lambda$是$\mathbb R$以上述等价关系为模的等价类的集合.<br />
这个集合是不可数的, 否则因为每个等价类是可数的, $\mathbb R$作为这些等价类的并集也将是可数的.</p>
<h2 id="几个引理">几个引理</h2>
<h3 id="引理一-测度函数的单调性">引理一 (测度函数的单调性)</h3>
<p>若$E \subseteq F$,则$\lambda (E) \le \lambda (F)$</p>

<p><strong>证明：</strong> <br />
$F = E \cup (F|E)$.<br />
由测度函数定义第四条，$\lambda (F) = \lambda (E \cup (F|E)) = \lambda (E) + \lambda (F|E)$。<br />
而$\lambda (F|E) \ge 0$,所以$\lambda (E) \le \lambda (F)$.</p>

<h3 id="引理二">引理二</h3>
<p>对$\forall p,q \in Q$,要么$\Omega  + p = \Omega  + q$,要么$(\Omega  + p) \cap (\Omega  + q) = \emptyset$.</p>
<blockquote>
  <p>也就是说, 两个这样的等价类要么相等, 要么不交</p>
</blockquote>

<p><strong>证明：</strong> <br />
反证, 假设$x\in (\Omega  + p) \cap (\Omega  + q)$, $p\neq q$, <br />
则存在$\alpha  \in \Omega$,使得$x = \alpha  + p$.同样，存在$\beta  \in \Omega$,使得$x = \beta  + q$.<br />
则$\alpha-\beta=q-p$.<br />
因为$p-q \in Q$,所以$\alpha-\beta \in Q$,由等价定义知，$\alpha \sim \beta$。<br />
再由$\Omega$的定义, $\alpha = \beta$, 那么$p = q$<br />
证毕。<br />
<img src="/assets/images/VataliSet.png" alt="示意图" /><br />
对这个证明可以这样理解:<br />
为了画图方便, 我在这里先假设$p=0$,$q \neq 0$. 那么$\Lambda$中的某个元素$x$向右平移$q$后,不可能有$\Lambda$中元素$y$和$x+q$重合(因为$\Lambda$中每个等价类只取一个, 而$x$和$x+q$是一个等价类的).</p>
<h2 id="不可测集的构造">不可测集的构造</h2>
<p>证明$\lambda$函数在Vatali集$\Omega$上的定义存在矛盾,即$\Omega$是不可测的.</p>

<p>因为 $\sum_{q \in \mathbb{Q},-1&lt;q&lt;1}(\Omega+q) \subseteq(-1,2)$</p>

<p>\(\lambda\left(\sum_{q \in \mathbb{Q},-1&lt;q&lt;1}(\Omega+q)\right) \leq \lambda((-1,2))=3\)<br />
再由可数可加性和平移不变性得<br />
\(\lambda\left(\sum_{q \in \mathbb{Q},-1&lt;q&lt;1}(\Omega+q)\right)=\sum_{q \in \mathbb{Q},-1&lt;q&lt;1} \lambda(\Omega+q)=\sum_{q \in \mathbb{Q},-1&lt;q&lt;1} \lambda(\Omega) \leq 3\)<br />
知<br />
\(\lambda(\Omega)=0\)<br />
故<br />
\(\sum_{q \in \mathbb{Q},-1&lt;q&lt;1} \lambda(\Omega)  = 0\)<br />
但是 $(0,1) \subseteq \sum_{q \in \mathbb{Q},-1&lt;q&lt;1}(\Omega+q) ， \lambda((0,1))=1$ ,</p>

<p>这样又可以得到<br />
\(\lambda\left(\sum_{q \in \mathbb{Q},-1&lt;q&lt;1}(\Omega+q)\right)=\sum_{q \in \mathbb{Q},-1&lt;q&lt;1} \lambda(\Omega+q)=\sum_{q \in \mathbb{Q},-1&lt;q&lt;1} \lambda(\Omega) \geq 1\)<br />
这是矛盾的.$\square$</p>

<p>解释一下 $(0,1) \subseteq \sum_{q \in \mathbb{Q},-1&lt;q&lt;1}(\Omega+q)$<br />
任取 $x \in(0,1)$ 則 $x=[x]+q,[x] \in \Omega \subseteq(0,1), q \in \mathbb{Q}$<br />
故 $q=x-[x] \in(-1,1)$ .</p>
<blockquote>
  <p>这里用[x]表示x所属的那个等价类的代表元</p>
</blockquote>]]></content><author><name>Bao Yu</name><email>baoyudu@outlook.com</email></author><summary type="html"><![CDATA[理想测度]]></summary></entry><entry><title type="html">线性空间</title><link href="https://baoyu.space/linear-space/" rel="alternate" type="text/html" title="线性空间" /><published>2021-01-18T15:00:00+00:00</published><updated>2021-01-18T15:00:00+00:00</updated><id>https://baoyu.space/linear-space</id><content type="html" xml:base="https://baoyu.space/linear-space/"><![CDATA[<h3 id="基本定义">基本定义</h3>

<blockquote class="def">
  <p>若对应法则$f: A\rightarrow B$满足每个$a\in A$对应唯一的$b\in B$, 则称 $\mathrm{f}$ 是 $\mathrm{A}$ 到 $\mathrm{B}$ 的一个映射, $\mathrm{b}$ 是 $\mathrm{a}$ 在 $\mathrm{f}$ 下的像, 记作 $\mathrm{f}(\mathrm{a})$. 称 $\mathrm{a}$ 是 $\mathrm{b}$ 在 $\mathrm{f}$ 下的一个原像.<br />
A称为<strong>定义域(Domain)</strong>,B称为<strong>陪域</strong>(Codomain).$f(A):= {f(a)\vert a\in A}$称为<strong>值域</strong>.</p>
</blockquote>

<blockquote class="def">
  <p>若$f(A) = B$,则f称为一个<strong>满射</strong>.<br />
如果$A$中不同元素在$f$下的像不同,称$f$为<strong>单射</strong>.<br />
如果$f$既是单射又是满射,则f称为一个<strong>双射</strong>(一一对应).</p>
</blockquote>
<hr />

<blockquote class="def">
  <p><strong>定义 笛卡尔积</strong></p>

  <p>$S \times M :={(a,b)|a \in S,b \in M} $ 称为$S$与$M$的<strong>笛卡尔积</strong>.</p>
</blockquote>

<hr />

<blockquote class="def">
  <p><strong>定义</strong></p>

  <p>非空集合$S$上的一个<strong>代数运算</strong>是指$S\times S$到$S$的一个映射.</p>

</blockquote>

<blockquote>
  <p>注意极限不是代数运算</p>
</blockquote>

<hr />

<blockquote class="def">
  <p><strong>定义</strong></p>

  <p>设$V$是一个非空集合,$K$是一个数域.<br />
如果$V$上有一个运算,称为加法,即</p>

\[(\alpha ,\beta ) \rightarrowtail \alpha+\beta\]

  <p>$K$和$V$之间有一个运算,称为<strong>数量乘法</strong>.即</p>

\[K \times V \rightarrow V:(k,\alpha) \rightarrowtail k\alpha\]

  <p>且满足8条运算性质<br />
V就称为数域K上的一个<strong>线性空间</strong>.</p>
</blockquote>

<h3 id="线性空间的例子n维向量空间">线性空间的例子:n维向量空间</h3>

<p>令<br />
\(K ^{n} = \{(a_1,a_2,...a_n)|a_i \in K,i = 1,2,...,n \}\)</p>

<p>规定<br />
 \((a_1,a_2,...a_n)=(b _1,b _2,...,b _n)\\
  \stackrel{\triangle}{\iff} \\
  a_i = b_i,i=1,2,...,n.\)</p>

<p>规定</p>

\[\begin{aligned}
    数量乘法\ &amp;k(a_1,a_2,,...,a_n)=(k a_1,k a_2 ,...,k a_n)\\
 加法\ &amp;(a_1,a_2,...a_n)+(b _1,b _2,...,b _n)=(a_1+b _1,a_2+b _2,...,a_n+b _n)
\end{aligned}\]

<p>并满足8条运算性质</p>

<table>
  <thead>
    <tr>
      <th>加法</th>
      <th>数乘</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>交换律</td>
      <td>单位元素</td>
    </tr>
    <tr>
      <td>结合律</td>
      <td>交换律</td>
    </tr>
    <tr>
      <td>零元素</td>
      <td>左分配律</td>
    </tr>
    <tr>
      <td>负元素</td>
      <td>右分配率</td>
    </tr>
  </tbody>
</table>

<p>则 $K^n是一个$K$上的$n$维向量空间.它是数域$K$上的一个线性空间.</p>

<h3 id="其他性质">其他性质</h3>

<p>设$V$$是数域$K$上的一个线性空间</p>

<blockquote class="thm">
  <p>V的零元唯一</p>
</blockquote>
<p><strong>证明</strong>:<br />
设 $O_1,O_2$都是V的零元,则<br />
\(O_1 = O_1+\stackrel{零元}{O_2}  =O_2+\stackrel{零元}{O_1}= O_2\)</p>

<blockquote class="thm">
  <p>每个$\alpha \in V$的负元唯一</p>
</blockquote>

<p><strong>证明</strong>:<br />
设$\beta_1,\beta_2$都是 $\alpha$的负元,则</p>

\[\beta _1 + \alpha +\beta _2=\beta _1+(\alpha + \beta _2)=(\beta_1+\alpha )+\beta _2 = \beta _1 = \beta _2\]

<blockquote class="thm">
  <p>$0\alpha = \stackrel{零元}{ \vec{0}}$</p>
</blockquote>

<p><strong>证明</strong>:<br />
\(0\alpha = (0+0)\alpha=0 \alpha + 0\alpha\)</p>

<p>两边加$0\alpha$的负元,得<br />
\(0 = 0 \alpha\)</p>

<blockquote class="thm">
  <p>$k\vec{0}= \vec{0}$</p>
</blockquote>

<p><strong>证明思路</strong>:<br />
零元是加法的性质,$k0$是数量乘法,要利用他们的”桥梁”性质$8$<br />
\(k0=k(0+0)=k0+k0\)</p>

<p>两边加$k0$的负元</p>

\[0 = k\vec{0}\]

<blockquote class="thm">
  <p>$k\alpha = 0$则k = 0或者$\alpha = 0$</p>
</blockquote>

<p><strong>证明</strong>:<br />
假设k = 0,则</p>

\[\alpha = 1\alpha = kk ^{-1} \alpha =k (k\alpha) = k\mathbf0 \xlongequal{由(4)} \mathbf{0}\]

<blockquote class="thm">
  <p>$(-1)\alpha = -\alpha$</p>
</blockquote>

<p><strong>证明</strong>:<br />
\((-1)\alpha + \alpha =(-1)\alpha + 1 \alpha =(-1+1)\alpha  = 0 \alpha \xlongequal{由(3)}\mathbf0\)</p>

<h3 id="线性子空间">线性子空间</h3>

<p>定义 设V是数域K上的一个线性空间,U是V的一个非空子集,如果U对于V的加法和数量乘法也成为数域K上的一个线性空间,那么称U是V的一个线性子空间.</p>]]></content><author><name>Bao</name></author><summary type="html"><![CDATA[基本定义]]></summary></entry><entry><title type="html">高等代数-二次型理论</title><link href="https://baoyu.space/ercixing/" rel="alternate" type="text/html" title="高等代数-二次型理论" /><published>2020-12-26T00:00:00+00:00</published><updated>2020-12-26T00:00:00+00:00</updated><id>https://baoyu.space/ercixing</id><content type="html" xml:base="https://baoyu.space/ercixing/"><![CDATA[<h3 id="二次型">二次型</h3>

<p>数域P上的一个n元二次型</p>

\[f(x_1,x_2,...,x_n)=\sum_{i=1}^n\sum_{j=1}^na_{ij}x_ix_j\]

<blockquote>
  <p>数域Ｐ上任意一个二次型都可以经过非退化的线性替换变成平方和的形式<br />
并称该平方和形式为该二次型的标准型</p>
</blockquote>

<blockquote>
  <p>证明</p>
</blockquote>

<p>设n为变量数</p>

<p>使用数学归纳法 $n=1时$</p>

\[f(x _{1}) = a _{11} ^{2}\]

<p>假设对n - 1 个变量的情形成立</p>

<p>n个变量时</p>

\[f(x _{1} ,x _{2} ,x _{3} , \cdots ,x _{n}  ) = \sum_{i = 1}^{n}\sum_{j=1}^{n}a _{ij} x _{i} x _{j}\]

<p>分三种情形来讨论</p>

<p><strong>Case 1</strong></p>

$a _{ii}$中至少有一个不为$0$，不妨设$a _{11}$ 不为零，这时
$$
\begin{align}
f(x _{1},x _{2},...,x _{3}) &= a _{11} x _{1} ^2 +\sum_{j=2}^{n}a _{1j}x _{1} x _{2} + \sum_{i=2}^{n}a _{i1} x _{i} x _{1} +\sum_{i= 2}^{n} \sum_{j=2}^{n} x _{i} x _{j}\\

&= a _{11} x _{1} ^2 +2\sum_{j=2}^{n}a _{1j}x _{1} x _{2}  +\sum_{i= 2}^{n} \sum_{j=2}^{n} x _{i} x _{j} \\   
&= a _{11} (x _{1} + \sum_{j=2}^{n}a _{11}^{-1}   a _{1j}x _{1} x _{j}    )^2 - (\sum_{i = 2}^{n}a _{1j}  x _{j} )^2 +\sum_{i = 2}^{n}\sum_{j= 2}^{n}a _{ij} x _{i} x _{j}
\end{align}
$$

<p>其中  $\sum_{i= 2}^{n} \sum_{j=2}^{n}a _{ij} x _{i}  x _{j}$<br />
是 $x _{2},x _{3} ,x _{4}$的二次型<br />
令</p>

\[\begin{cases}
    y_1 = x _{1} + \sum_{j =2}^{n}a _{11} a _{1j}x _{j}
    &amp;\text{}\\
    y _{2} = x_2 \\
    \vdots \\
    y_n = x_n
     &amp;\text{}\\
\end{cases} \\\]

<p>即</p>

\[\begin{cases}
    x_1 = y _{1} - \sum_{j =2}^{n}a _{11} a _{1j}x _{j}
    &amp;\text{}\\
    x _{2} = y_2 \\
    \vdots \\
    x_n = y_n
     &amp;\text{}\\
\end{cases} \\\]

<p>这是一个非退化线性替换</p>

\[s.t. f(x_1,x_2,...,x_n)=a _{11}y _{1} ^2+\sum_{i=2}^{n}b _{ij} y_iy_j.\]

<p>根据MI假设，对 $  \sum_{i=2}^{n}b _{ij} y_iy_j $, 有非退化线性替换</p>

\[\begin{pmatrix}
z_1\\
z_2\\
...\\
z_n\\
\end{pmatrix}=\begin{pmatrix}
c_{11}&amp;c_{12}&amp;\cdots&amp;c_{14}	\\
c_{21}&amp;c_{22}&amp;\cdots&amp;c_{24}	\\
\vdots&amp;\vdots&amp;&amp;\vdots	\\
c_{41}&amp;c_{42}&amp;\cdots&amp;c_{44}	\\
\end{pmatrix}\\
s.t.\\ f(x_1,x_2,...,x_n) = a _{11} z_1 ^2 + z_2 ^2 + ...+z_n ^2\]

<p><strong>Case 2</strong></p>

<p>$\forall i,a _{ii} = 0$,but  $\exists i,s.t. \ a _{1j}=0$</p>

<p>不失一般性，设 $a _{12} \neq 0$<br />
令</p>

\[\begin{cases}
  x_1 = z_1 + z_2\\
  x_2 = z_1 - z_2\\
  x_3 = z_3\\
  ...\\
  x_n= z_n 
\end{cases}\]

<p>他是非退化线性替换，且使</p>

\[f(x_1,x_2,..,x_n)= 2a _{12}x_1 x_2 +...\\
 =2a _{12} (z_1 + z_2)(z_1-z_2)+...\\
 =2 a _{12}z_1 ^2-2a _{12}  z_2 ^2\]

<p>且 $ z_1 ^2 $的系数不为0，属于第一种情况，成立.</p>

<p><strong>Case 3</strong></p>

<p>$a _{11} =a _{12}=…=a _{1n}=0$</p>

<p>由对称性，</p>

<p>$a_{11} = a _{21}=…=a _{n1}$</p>

<p>为 $x_2,x_3,…,x_n的二次型$</p>

<p>由归纳假设，可表示为平方和的形式.</p>

<p>至此原命题证毕 #</p>

<h3 id="二次型的矩阵">二次型的矩阵</h3>

\[A = \begin{pmatrix}
a_{11}&amp;a_{12}&amp;\cdots&amp;a_{14}	\\
a_{21}&amp;a_{22}&amp;\cdots&amp;a_{24}	\\
\vdots&amp;\vdots&amp;&amp;\vdots	\\
a_{41}&amp;a_{42}&amp;\cdots&amp;a_{44}	\\
\end{pmatrix}\\
f(x_1,x_2,...,x_n)= \begin{pmatrix}
x_1&amp;x_2&amp;...&amp;x_n
\end{pmatrix}
\begin{pmatrix}
a_{11}&amp;a_{12}&amp;\cdots&amp;a_{14}	\\
a_{21}&amp;a_{22}&amp;\cdots&amp;a_{24}	\\
\vdots&amp;\vdots&amp;&amp;\vdots	\\
a_{41}&amp;a_{42}&amp;\cdots&amp;a_{44}	\\
\end{pmatrix}\begin{pmatrix}
x_1\\
x_2\\
\vdots \\
x_n
\end{pmatrix}\\
= C ^{'}AC\]

<blockquote>
  <p>定理2<br />
在数域P上，任何一个对称矩阵都合同于一个对角矩阵，并称</p>
</blockquote>

<p>证明：</p>

<p>将上述证明以矩阵形式写出即可.</p>

<p>使用同前面证明类似的数学归纳法</p>

<p>$<br />
1.a _{ii}不全为0时 <br />
$</p>

<p>不失一般性，设 $a _{11} \neq 0  $</p>

<p>应用变换<br />
\(C_1 = \begin{pmatrix}
1&amp; -a _{11}^{-1}  a _{12} &amp; ...&amp;-a _{11}^{-1}a _{1n}\\
0&amp;1&amp;...&amp;0\\
\vdots &amp;\vdots &amp; &amp;\vdots\\
0&amp;0&amp;...&amp;1
\end{pmatrix}\)</p>

<p>合同变换为</p>

<p>$C^{‘}AC$<br />
令<br />
\(\alpha = \begin{pmatrix}
a _{12} \cdots  a _{1n} 
 \end{pmatrix}\\
 A_1 = \begin{pmatrix}
 a_{22}&amp;\cdots&amp;a_{24}	\\
\vdots&amp;&amp;\vdots	\\
a_{42}&amp;\cdots&amp;a_{44}	\\
 \end{pmatrix}\)</p>

<p>于是C,A,C’都可以写成分块形式：</p>

<p>\(A = \begin{pmatrix}
 a _{11} &amp;\alpha \\ 
 \alpha ' &amp;A_1
 \end{pmatrix} 
C =\begin{pmatrix}
1&amp;-a _{11} ^{-1}  \alpha \\
0&amp;E_n
\end{pmatrix}\)<br />
则</p>

\[\begin{array}
 \\C'AC = \\
\begin{pmatrix}
1&amp;0\\
-a _{11} ^{-1} &amp;E _{n}  
\end{pmatrix}\begin{pmatrix}
a _{11} &amp; \alpha \\
\alpha '&amp;A _{1} 
\end{pmatrix}\begin{pmatrix}
1&amp;-a _{11} \alpha \\
0&amp;E_n
\end{pmatrix}
 \end{array}\\=\begin{pmatrix}
1&amp;0\\
-a _{11} ^{-1} &amp;E _{n}  
\end{pmatrix}\begin{pmatrix}
a _{11} &amp;0\\\alpha '&amp;A_1-a _{11}^{-1}  \alpha \alpha '
\end{pmatrix}\\=\begin{pmatrix}a _{11} &amp; 0\\
0&amp; A_1 -a _{11} ^{-1} \alpha \alpha '
\end{pmatrix}\]

<p>$A_1-a _{11}^{-1}  \alpha \alpha ‘$ 是一个n-1级对称方矩阵，</p>

<p>由归纳假设，可由可逆矩阵G使<br />
\(G'(A_1-a _{11}^{-1}  \alpha \alpha ')G = D\)</p>

<p>为对角矩阵</p>

<p>令<br />
\(C_2 = \begin{pmatrix}
 1&amp;0\\
 0&amp;G  
 \end{pmatrix} \\
 C_2'C_1'AC_1C_2=\begin{pmatrix}
 1&amp;0\\
 0&amp;G'  
 \end{pmatrix}\begin{pmatrix}a _{11} &amp; 0\\
0&amp; A_1 -a _{11} ^{-1} \alpha \alpha '
\end{pmatrix}\begin{pmatrix}
 1&amp;0\\
 0&amp;G'  
 \end{pmatrix}\\=\begin{pmatrix}
 1&amp;0\\0&amp;D
 \end{pmatrix}为对角矩阵\\
 C_1C_2即为所求可逆矩阵\)<br />
2.<br />
$a _{ii} \equiv  0 , \forall  i,<br />
\exists a _{1i} \neq 0$</p>

<p>应用变换</p>

\[C = \begin{pmatrix}
   1 &amp; -1 &amp;0&amp;\cdots&amp; 0 	\\
   1 &amp; 1 &amp;0&amp;\cdots&amp; 0 \\
   0&amp;0&amp;1&amp;...&amp;0	\\
  \vdots&amp;\vdots&amp;\vdots &amp;&amp;\vdots	\\
   0 &amp; 0 &amp;0&amp;\cdots&amp; 1 	\\
  \end{pmatrix}\]

<p>则</p>

\[C'AC = \begin{pmatrix}
 2a _{12}&amp;0&amp;...\\
 0&amp;-2a _{12}&amp;...\\
 ...&amp;...&amp;...  
 \end{pmatrix}\]

<p>转化为第一种情况.</p>

<p>3.$\forall i ,a _{1i} = a _{i1}=0$</p>

<p>矩阵写为</p>

<p>\(\begin{pmatrix}
 0&amp;0\\
 0&amp;A_1
 \end{pmatrix}\)<br />
其中$A_1$为n-1个变量的对称矩阵，由归纳假设，与对角矩阵合同.</p>

<p>至此原命题证毕.#</p>

<h3 id="唯一性">唯一性</h3>
<p>在一般的数域内，二次型的标准形不是唯一的，而与所作的非退化线性替换有关.</p>

<h5 id="复数域情形">复数域情形</h5>

<p>$f(x_1,x_2,…,x_n)$是复数域内的一个二次型,由上述定理1,可经可逆线性替换变为标准型</p>

<p>不妨假设为<br />
\(d_1 y_1 ^2 d_2 y_2 ^2 + ... + d_r y_r ^2 \\
 d_i \neq 0,i = 1,2,...,r.\)</p>

<p>r就是二次型的秩</p>

<p>再通过可逆线性替换(注意复数总可以开平方)<br />
\(\begin{cases}
    y_1 = \frac{1}{\sqrt{d_1}}z_1 &amp;\text{}\\
    y_2 = \frac{1}{\sqrt{d_2}}z_2 &amp;\text{}\\
    ...\\
    y_r = \frac{1}{\sqrt{d_r}}z_n\\
    y _{r+1} = z _{r+1} \\
    ...\\
    y_n = z_n
  \end{cases}\)<br />
即得<br />
\(z_1 ^2+z_2 ^2 + ...+z_n ^2\)</p>

<p>称为复二次型$f(x_1,…,x_n)$的规范形,显然,在复数域内,规范性完全由原二次型矩阵的秩决定</p>

<blockquote>
  <p>定理3<br />
任意一个复二次型可经非退化线性变换化为唯一确定的规范形.</p>
</blockquote>

<blockquote>
  <p>或者说,任意复数域内的对称矩阵合同于一形如 $ \begin{pmatrix}<br />
1&amp;0&amp;…\<br />
0&amp;1&amp;…\<br />
0&amp;0&amp;…<br />
\end{pmatrix} $的对角矩阵.</p>
</blockquote>

<blockquote>
  <p>推论<br />
两个复数矩阵合同的充要条件是他们的秩相等</p>
</blockquote>

<p>####实数域情形</p>

<p>$f(x_1,x_2,…,x_n)$是一实二次型,由定理一,经非退化线性替换,再适当排列文字的次序,得<br />
\(d _{1} y_1 ^2+d_2 y_2 ^2 + ...+d_p y_p - d_{p+1}- d _{p+1} ^2  ... - d _r y_r ^2\)<br />
其中 $d_i &gt; 0 ,i = 1,2,…,r ,r = rank(A)$</p>

<p>\(\begin{cases}
    y_1 = \frac{1}{\sqrt{d_1}}z_1 &amp;\text{}\\
    y_2 = \frac{1}{\sqrt{d_2}}z_2 &amp;\text{}\\
    ...\\
    y_r = \frac{1}{\sqrt{d_r}}z_n\\
    y _{r+1} = z _{r+1} \\
    ...\\
    y_n = z_n
  \end{cases}\)<br />
就得到<br />
\(z_1 ^2 + z_2 ^2 + ... + z_p ^2 -z _{p+1} ^2 - ... - z_r ^2\)<br />
称为实二次型 $f(x_1,x_2,…x_n)$的规范形,显然,它完全被r,p两个数决定.</p>

<blockquote>
  <p>定理4<br />
任意一个实二次型可经过适当的非退化线性替换可化为唯一确定的规范形</p>
</blockquote>

<p>前半部分(可化为规范型)已经证明,下证唯一性</p>

<p>设实二次型$f(x_1,x_2,…,x_n)$经非退化的线性替换</p>

\[X= BY\]

<p>化为规范型</p>

\[y_1 ^2+ y_2 ^2 + \cdots + y_p^2- \cdots - y_r ^2\]

<p>经非退化线性替换</p>

\[X = CZ\]

<p>化为规范型</p>

\[z_1 ^2 +z_2 ^2 + \cdots  z_p ^2 - \cdots - z_r ^2\]

<p>用反证法.设p&gt;q.</p>

<p>由以上假设,我们有</p>

\[y_1 ^2+ y_2 ^2 + \cdots + y_p^2- \cdots - y_r ^2 = z_1 ^2 +z_2 ^2 + \cdots  z_p ^2 - \cdots - z_r ^2 \tag{*}\]

<p>其中 $Z = C^{-1}BY\tag{1}$</p>

<p>令<br />
\(\begin{pmatrix}
g_{11}&amp;g_{12}&amp;\cdots&amp;g_{1n}	\\
g_{21}&amp;g_{22}&amp;\cdots&amp;g_{2n}	\\
\vdots&amp;\vdots&amp;&amp;\vdots	\\
g_{n1}&amp;g_{n2}&amp;\cdots&amp;g_{nn}	\\
\end{pmatrix}\)</p>

<p>所以(1)可以写成</p>

\[\begin{cases}
  z_1 = g_{11} y_1 + g_{12}y_2+ \cdots +g_{1n}y_n ,\\
   z_2 = g_{21} y_2 + g_{22}y_2+ \cdots +g_{2n}y_n, \\
   \cdots\\
   z_n = g_{n1}y_1+ g_{n2}y_2+ \cdots +g_{nn}y_n
\end{cases} \tag{2}\]

<p>考虑齐次线性方程组</p>

<blockquote>
  <p>这里的思路来自北大《高等代数》(第四版 北京大学前代数小组 高等教育出版社)个人认为一定程度上缺少直觉性.</p>
</blockquote>

\[\begin{cases}
  g_{11}y_1+g_{12}y_2+ \cdots+g_{1n}y_n = 0,\\
  \cdots \\
    g_{q1}y_1+g_{q2}y_2+ \cdots+g_{qn}y_n = 0,\\
    y _{p+1} = 0,\\
    \cdots \\
    y_n=0.
\end{cases} \tag{3}\]

<p>方程组(3)含有n个未知量,$q+(n-p)=n-(p-q)&lt;n$个方程,故方程组有非零解.令</p>

\[(y_1,...,y_p,y_{p+1},...,y_n) =(k _1,...,k_p,k_{p+1},...,k _n)\]

<p>是(3)的一个非零解.显然</p>

\[k_{p+1}= ...=k _n= 0\]

<p>因此,把它带入(*)的左端,得到的值为</p>

\[k _1 ^2+ k _2 ^2 +...+k _{p} &gt; 0\]

<p>再通过(2)把它带入(*)的右端,因为它是(3)的解,故有</p>

\[z_1=z_2=...=z_q=0\]

<p>所以</p>

\[-z _{q+1} ^2 -z _{q+2} -...-z _{r} ^2  \leqslant 0\]

<p>矛盾.$p\leq q$得证,同理可证$p\geq q$.<br />
于是唯一性得证.<br />
这个定理通常称为 <strong>惯性定理</strong>.</p>]]></content><author><name>Bao</name></author><summary type="html"><![CDATA[二次型]]></summary></entry></feed>