【CQF Math Class 数学笔记】

CQF Math Class Notes

Bayes’ Probability

$$P(B|A)=\frac{P(A|B)\ast P(B)}{P(A)}$$
or
$$P(B|A)\ast P(A)=P(A\cap B)=P(B\cap A)=P(A|B)\ast P(B)$$
我们的假设条件是A事件基于B的发生的概率与B事件基于A发生的概率是一样的。
举个例子：从一副牌连续翻出两张K的概率是多少 (大小鬼被剔除)？
$$K\cap K=4/52\ast 3/51=1/221\approx 0.5\%$$
我们从上面的例子可以看出，两个条件互换所得出的结果是一样的。从而得出上面贝叶斯概率公式的结论。

---

The Crazy Intergral and its Rules and Formulas

我们如何来证明这道题？可以说我们基本上把能用的微积分知识和定理都撸了一遍，这是一道非常好的一个练习题：
$$\int \frac{1}{(x^2+1)}dx=ta{n}^{-1}x+C$$
$$=\int \frac{1}{(x+i)(x-i)}dx=\int \frac{(1/2i)}{(x+i)}dx+\int \frac{(-i/2i)}{(x-i)}dx$$
$$=\frac{1}{2i}(ln(x-i)-ln(x+i))$$
这个答案看着好像已经结束了，当时这是不可以的也不应该！因为我们原问题并不包含虚数，所以答案里包含虚数是不合理的，所以要继续简化！
下面我们要用到欧拉恒等式（Euler’s Equation）$e^{i\theta }=\cos \theta +i\sin \theta$来进行运算：
$a+bi=r{e}^{i\theta }$
$r=\sqrt{(a^2+b^2)}$
$\theta =ta{n}^{-1}\frac{b}{a}$
然后我们继续进行替换：
$$=\frac{1}{2i}(ln(x-i)-ln(x+i))=\frac{1}{2i}(ln(\sqrt{(x^2+(-1{)}^2)}\ast e^{ita{n}^{-1}\frac{-1}{x}})-ln(\sqrt{(x^2+(1{)}^2)}\ast e^{ita{n}^{-1}\frac{1}{x}}))$$
$$=\frac{1}{2i}(ln\sqrt{(x^2+1)}+ln(e^{ita{n}^{-1}\frac{-1}{x}})-ln\sqrt{(x^2+1)}-ln(e^{ita{n}^{-1}\frac{1}{x}}))$$
$ln\sqrt{(x^2+1)}抵消掉$，$ln(e)抵消掉$，
$ta{n}^{-1}$是一个奇函数（odd function），所以我们可以将(-1/x)的提出负号出来，即：
$$=\frac{1}{2i}(-2i\ast ta{n}^{-1}\frac{1}{x})=-ta{n}^{-1}\frac{1}{x}+C1$$
其实做到这里我们可以算是做完了，我个人觉得老师未必会扣你的分数。但是呢，其实还是可以有一丢丢的优化空间，因为：
$ta{n}^{-1}x+co{t}^{-1}x=\frac{\pi }{2}$，且 ${\cot }^{-1}x={\tan }^{-1}\frac{1}{x}$，左右移动项得出：
$$={\tan }^{-1}x-\frac{\pi }{2}+C1$$
$\frac{\pi }{2}+C1$他们其实都可以看作是常数项，所以可以合并，最终最优解，即：
$$\int \frac{1}{(x^2+1)}dx=ta{n}^{-1}x+C$$

---

Trignometry Formula

$\cos (\alpha +\beta )=\cos \alpha \ast \cos \beta -\sin \alpha \ast \sin \beta$
$\cos (\alpha -\beta )=\cos \alpha \ast \cos \beta +\sin \alpha \ast \sin \beta$
$\sin (\alpha \pm \beta )=\sin \alpha \ast \cos \beta \pm \cos \alpha \ast \sin \beta$

$\tan (\alpha +\beta )=\frac{\tan \alpha +\tan \beta }{1-\tan \alpha \ast \tan \beta }$

$\tan (\alpha -\beta )=\frac{\tan \alpha -\tan \beta }{1-\tan \alpha \ast \tan \beta }$

---

Forward Kolmogorov Equation

The equation is given by

$\frac{\delta p}{\delta t^{\prime }}=C^2\frac{{\delta }^{2}p}{\delta y^{\prime 2}}$

for the transition density function $p(y,t;y^{\prime },t^{\prime });c^2\in \mathbb{R}$, The states (y,t) are past and are fixed while (y’,t’) refers t ofuture ones and are variables. By simple substitution show that

$p(y,t;y^{\prime },t^{\prime })=\frac{1}{2c\sqrt{\pi (t^{\prime }-t)}}exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})$ ---- (1.2) satisfies the KFE

show that 1.2.satisfies
${\int }_{-\infty }^{\infty }p(y,t;y^{\prime },t^{\prime })d{y}^{\prime }=1$

let’s solve this problem.

LHS:

$\frac{\delta p}{\delta t^{\prime }}=\frac{1}{2c\sqrt{\pi }}\ast exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})\ast -\frac{1}{2}(t^{\prime }-t{)}^{\frac{3}{2}}+\frac{1}{2c\sqrt{\pi (t^{\prime }-t)}}exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})\ast \frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t{)}^2}$ ---- product rule

$=\frac{(y^{\prime }-y{)}^2-2c^2(t^{\prime }-t)}{8c^3\sqrt{\pi }(t^{\prime }-t{)}^{\frac{5}{2}}}exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})$

RHS:

$\frac{\delta p}{\delta y^{\prime }}=\frac{1}{2c\sqrt{\pi (t^{\prime }-t)}}\ast exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})\ast \frac{-2(y^{\prime }-y)}{4c^2(t^{\prime }-t)}$

$=-\frac{(y^{\prime }-y)}{(4c^3\sqrt{\pi }(t^{\prime }-t{)}^{\frac{3}{2}}}\ast exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})$

$\frac{{\delta }^{2}p}{\delta y^{\prime 2}}=-\frac{1}{(4c^3\sqrt{\pi }(t^{\prime }-t{)}^{\frac{3}{2}}}\ast exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})-\frac{(y^{\prime }-y)}{(4c^3\sqrt{\pi }(t^{\prime }-t{)}^{\frac{3}{2}}}\ast exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})\ast \frac{-2(y^{\prime }-y)}{4c^2(t^{\prime }-t)}$

$=\frac{(y^{\prime }-y{)}^2-2c^2(t^{\prime }-t)}{(8c^5\sqrt{\pi }(t^{\prime }-t{)}^{\frac{5}{2}}}\ast exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})$

Hence: $LHS=\frac{\delta p}{\delta t^{\prime }}=\frac{(y^{\prime }-y{)}^2-2c^2(t^{\prime }-t)}{8c^3\sqrt{\pi }(t^{\prime }-t{)}^{\frac{5}{2}}}exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})=C^2\frac{{\delta }^{2}p}{\delta y^{\prime 2}}=RHS$

we know that $\frac{(y^{\prime }-y{)}^2-2c^2(t^{\prime }-t)}{8c^3\sqrt{\pi }(t^{\prime }-t{)}^{\frac{5}{2}}}exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})$ is a common solution to FKE, then let’s see if its PDF is equal to 1 as all PDF should sum up to 1.

${\int }_{-\infty }^{\infty }p(y,t;y^{\prime },t^{\prime })d{y}^{\prime }={\int }_{-\infty }^{\infty }\frac{1}{2c\sqrt{\pi (t^{\prime }-t)}}exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})d{y}^{\prime }$

from observation, it is better to use chain rule in this situation, and it’ll make things a lot easier.

let $u=\frac{(y^{\prime }-y)}{2c\sqrt{(}{t}^{\prime }-t)}$ then $\frac{du}{d{y}^{\prime }}=\frac{1}{2c\sqrt{(t^{\prime }-t)}}$, thus $d{y}^{\prime }=2c\sqrt{(t^{\prime }-t)}du$

${\int }_{-\infty }^{\infty }\frac{1}{2c\sqrt{\pi (t^{\prime }-t)}}exp(-\frac{(y^{\prime }-y{)}^2}{4c^2(t^{\prime }-t)})d{y}^{\prime }\Rightarrow \frac{1}{2c\sqrt{\pi (t^{\prime }-t)}}\ast {\int }_{-\infty }^{\infty }{e}^{-u^2}du\ast 2c\sqrt{(t^{\prime }-t)}=\frac{1}{\sqrt{\pi }}\ast {\int }_{-\infty }^{\infty }{e}^{-u^2}du=1$

why? because ${\int }_{-\infty }^{\infty }{e}^{-u^2}du=\sqrt{\pi }$ is known by definition. and we are done with this quesiton.

— 待更 —