CQF笔记Primer数学基础
- 1 Calculus 微积分
- 2 Linear Algebra 线性代数
- 3 Differential Equations 微分方程
- 3.1 Introduction 简介
- 3.2 First Order Ordinary Differential Equations 一阶常微分方程
- 3.3 Second Order ODE.s 二阶ODE
- 3.4 General nth Order Equation n阶方程
- 3.5 Non-Homogeneous Case - Method of Undetermined Coefficients 非齐次
- 3.6 Linear ODE.s with Variable Coefficients - Euler Equation 可变系数的线性ODE - (柯西-)欧拉方程
- 3.7 Partial Differential Equations 偏微分方程(略)
- 1 Probability 概率
- 1.1 Preliminaries 知识准备
- 1.2 Probability Diagrams 用图表表示概率
- 1.3 Conditional Probability 条件概率
- 1.4 Mutually exclusive and Independent events 互斥和独立
- 1.5 Two famous problems
- 1.6 Random Variables 随机变量
- 1.7 Probability Distributions 概率分布
- 1.8 Cumulative Distribution Function CDF 累积分布函数
- 1.9 Expectation and Variance 期望和方差
- 1.10 Expectation Algebra 期望值的代数运算
- 1.11 Moments 矩
- 1.12 Covariance 协方差
- 1.13 Important Distributions 重要分布
- 1.14 Central Limit Theorem 中心极限定理
- 2 Statistics 统计
1 Calculus 微积分
1.1 Basic Terminology 基本术语
符号 | 含义 | 符号 | 含义 | 符号 | 含义 |
---|---|---|---|---|---|
∃ \exists ∃ | 存在 | → \rightarrow → | 给定 | ≡ \equiv ≡ | 等价于 |
∀ \forall ∀ | 对所有……有 | s.t. | such that | ∼ \sim ∼ | similar |
∴ \therefore ∴ | 所以 | : | such that | ∈ \in ∈ | 是……的元素 |
∵ \because ∵ | 因为 | iff | 当且仅当 | !x | 有唯一的x |
1.2 Functions 函数
输入到输出的映射 mapping:
- input记作x,independent variable,自变量
- output记作y,dependent variable,因变量
mapping 映射:
- 一对一
- 多对一
- 一对多(不是函数)
函数的定义:每个x映射到唯一的一个y
inverse function 逆函数
- 一对一映射的函数有反函数
- 多对一映射的函数没有反函数,限制自变量的值域,变成一对一映射
函数
y
=
2
x
2
−
1
y = 2x^2 - 1
y=2x2−1的逆函数(限制x的取值范围为
x
≥
0
x \ge 0
x≥0)为
y
=
x
+
1
2
y = \sqrt{\frac{x+1}{2}}
y=2x+1
f
(
f
−
1
(
x
)
)
=
x
f(f^{-1}(x)) = x
f(f−1(x))=x
f
−
1
(
f
(
x
)
)
=
x
f^{-1}(f(x)) = x
f−1(f(x))=x
even function偶函数和 odd function 奇函数
偶函数
- f ( − x ) = f ( x ) f(-x) = f(x) f(−x)=f(x)
- 沿y轴对称
奇函数
- f ( − x ) = f ( − x ) f(-x) = f(-x) f(−x)=f(−x)
- 中心对称(沿原点旋转180°)
大部分函数不是偶函数或者奇函数,但是可以表达为奇函数和偶函数之和
1.2.1 Explicit/Implicit Representation 显示/隐式表示
显示表示
y
=
2
x
2
+
4
x
−
16
=
0
y = 2x^2 + 4x - 16 = 0
y=2x2+4x−16=0
隐式表示
f
(
x
,
y
)
=
0
f(x,y) = 0
f(x,y)=0
1.2.2 Types of function
Polynomials 多项式
f ( x ) = ∑ k = 0 n a k x k , 其 中 a 0 , a 1 , . . . , a n 是 常 数 f(x) = \sum_{k=0}^{n} a_k x^k, \ 其中a_0, a_1, ..., a_n是常数 f(x)=k=0∑nakxk, 其中a0,a1,...,an是常数
- K = 1称为线性项
- K = 2称为二次项
二次多项式
a
x
2
+
b
x
+
c
=
0
ax^2 + bx + c = 0
ax2+bx+c=0 的解
x
=
−
b
±
b
2
−
4
a
c
2
a
x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}
x=2a−b±b2−4ac
三种情况:
- b 2 − 4 a c > 0 → x 1 ≠ x 2 , 两 个 实 数 根 b^2 - 4ac > 0 \rightarrow x_1 \neq x_2,两个实数根 b2−4ac>0→x1=x2,两个实数根
- b 2 − 4 a c = 0 → x 1 = x 2 = − b 2 a 为 实 数 b^2 - 4ac = 0 \rightarrow x_1 = x_2 = - \frac{b}{2a} 为实数 b2−4ac=0→x1=x2=−2ab为实数
- b 2 − 4 a c < 0 → x 1 ≠ x 2 , 两 个 复 数 共 轭 根 b^2 - 4ac < 0 \rightarrow x_1 \neq x_2,两个复数共轭根 b2−4ac<0→x1=x2,两个复数共轭根
Modulus 模(绝对值)
f ( x ) = { x x>0 − x x<0 f(x)= \begin{cases} x& \text{x>0}\\ -x& \text{x<0} \end{cases} f(x)={x−xx>0x<0
模函数是分段函数piecewise function
1.3 Limit 极限
研究逼近问题
lim x → x 0 f ( x ) → l \lim_{x \to x_0} f(x) \rightarrow l x→x0limf(x)→l
同时从左边逼近和从右边逼近,应该相同逼近同一个值
普通极限
lim x → ∞ x 2 + 2 x + 2 3 x 2 + 4 = lim x → ∞ 1 + 2 x + 2 x 2 3 + 4 x 2 = 1 3 \begin{aligned} & \lim_{x \to \infty} \frac{x^2 + 2x + 2}{3x^2 + 4} \\ = & \lim_{x \to \infty} \frac{1 + \frac{2}{x} + \frac{2}{x^2}}{3 + \frac{4}{x^2}} \\ = & \frac{1}{3} \end{aligned} ==x→∞lim3x2+4x2+2x+2x→∞lim3+x241+x2+x2231
函数连续
lim x → x 0 f ( x ) = f ( x 0 ) \lim_{x \to x_0} f(x) = f(x_0) x→x0limf(x)=f(x0)
1.3.1 exponential and log functions 指数和对数函数
y = a x y = log a x y = a^x \\ y = \log_a x y=axy=logax
以自然对数为底的指数函数和对数函数
y
=
e
x
y
=
l
n
x
y
=
l
o
g
x
y = e^x \\ y = ln\ x \\ y = log\ x
y=exy=ln xy=log x
e = lim n → ∞ ( 1 + x n ) n e = \lim_{n \to \infty} (1 + \frac{x}{n})^n e=n→∞lim(1+nx)n
正态分布的形式类似于 e − x 2 / 2 e^{-x^2 / 2} e−x2/2
1.3.2 Trigonometric/Circular Functions 三角函数
sin 正弦函数
- 奇函数 s i n ( − x ) = − s i n ( x ) sin(-x) = -sin(x) sin(−x)=−sin(x)
- 周期函数 s i n ( x + 2 π ) = s i n ( x ) sin(x + 2 \pi) = sin(x) sin(x+2π)=sin(x)
- s i n x = 0 ⇔ x = n π ∀ n ∈ Z sinx = 0 \Leftrightarrow x = n \pi \ \forall n \in Z sinx=0⇔x=nπ ∀n∈Z
- D o m ( s i n x ) = R Dom(sinx) = R Dom(sinx)=R and I m ( s i n x ) = [ − 1 , 1 ] Im(sinx) = [-1, 1] Im(sinx)=[−1,1]
cos 余弦函数
- 偶函数 c o s ( − x ) = c o s ( x ) cos(-x) = cos(x) cos(−x)=cos(x)
- 周期函数 c o s ( x + 2 π ) = c o s ( x ) cos(x + 2 \pi) = cos(x) cos(x+2π)=cos(x)
- c o s = 0 ⇔ x = ( 2 n + 1 ) π 2 ∀ n ∈ Z cos= 0 \Leftrightarrow x = (2n + 1) \frac{\pi}{2} \ \forall n \in Z cos=0⇔x=(2n+1)2π ∀n∈Z
- D o m ( c o s x ) = R Dom(cos\ x) = R Dom(cos x)=R and I m ( c o s x ) = [ − 1 , 1 ] Im(cos \ x) = [-1, 1] Im(cos x)=[−1,1]
tan 正切函数
- 偶函数 t a n ( − x ) = t a n ( x ) tan (-x) = tan (x) tan(−x)=tan(x)
- 周期函数 c o s ( x + π ) = c o s ( x ) cos(x + \pi) = cos(x) cos(x+π)=cos(x)
- D o m ( t a n x ) = { x : c o s x ≠ 0 } = { x : x ≠ ( 2 n + 1 ) π 2 ; ( n ∈ Z ) } = R − { x = ( 2 n + 1 ) π 2 ; ( n ∈ Z ) } Dom(tan\ x) = \{x : cos \ x \neq 0\} = \{x : x \neq (2n + 1) \frac{\pi}{2}; \ (n \in Z) \} = R - \{x = (2n + 1) \frac{\pi}{2}; \ (n \in Z) \} Dom(tan x)={x:cos x=0}={x:x=(2n+1)2π; (n∈Z)}=R−{x=(2n+1)2π; (n∈Z)}
三角函数关系式
c
o
s
2
x
+
s
i
n
2
x
=
1
cos^2 x + sin^2 x = 1 \\
cos2x+sin2x=1
s i n ( x ± y ) = s i n x c o s y ∓ c o s x s i n y c o s ( x ± y ) = c o s x c o s y ∓ s i n x s i n y t a n ( x ± y ) = t a n x ± t a n y 1 ∓ t a n x t a n y sin(x \pm y) = sinx \ cosy \mp cosx \ siny \\ cos(x \pm y) = cosx \ cosy \mp sinx \ siny \\ tan(x \pm y) = \frac{tanx \pm tany}{1 \mp tanx \ tany} \\ sin(x±y)=sinx cosy∓cosx sinycos(x±y)=cosx cosy∓sinx sinytan(x±y)=1∓tanx tanytanx±tany
s e c x = 1 c o s x c s c x = 1 s i n x c o t x = 1 t a n x secx = \frac {1} {cosx} \\ cscx = \frac {1} {sinx} \\ cotx = \frac {1} {tanx} \\ secx=cosx1cscx=sinx1cotx=tanx1
s i n − 1 x → a r c s i n ( x ) c o s − 1 x → a r c c o s ( x ) t a n − 1 x → a r c t a n ( x ) sin^{-1}x \rightarrow arcsin(x) \\ cos^{-1}x \rightarrow arccos(x) \\ tan^{-1}x \rightarrow arctan(x) \\ sin−1x→arcsin(x)cos−1x→arccos(x)tan−1x→arctan(x)
lim x → 0 s i n x = 0 lim x → 0 s i n x x = 1 lim x → 0 ∣ x ∣ = 0 lim x → 0 ∣ x ∣ x 的 左 右 极 限 分 别 为 − 1 和 1 , 因 此 没 有 极 限 \lim_{x \to 0} sinx = 0 \\ \lim_{x \to 0} \frac{sinx}{x} = 1 \\ \lim_{x \to 0} |x| = 0 \\ \lim_{x \to 0} \frac{|x|}{x} 的左右极限分别为-1和1,因此没有极限 \\ x→0limsinx=0x→0limxsinx=1x→0lim∣x∣=0x→0limx∣x∣的左右极限分别为−1和1,因此没有极限
1.3.3 Hyperbolic Functions 双曲函数
s i n h x = 1 2 ( e x − e − x ) sinh \ x = \frac{1}{2}(e^x-e^{-x}) sinh x=21(ex−e−x)
- 奇函数
- D o m ( s i n h x ) = R Dom(sinh \ x) = R Dom(sinh x)=R and I m ( s i n h x ) = R Im(sinh \ x) = R Im(sinh x)=R
c o s h x = 1 2 ( e x + e − x ) cosh \ x = \frac{1}{2}(e^x+e^{-x}) cosh x=21(ex+e−x)
- 偶函数
- D o m ( c o s h x ) = R Dom(cosh\ x) = R Dom(cosh x)=R and I m ( c o s h x ) = [ 1 , ∞ ) Im(cosh\ x) = [1, \infty) Im(cosh x)=[1,∞)
t a n h x = s i n h x c o s h x tanh \ x = \frac{sinhx}{coshx} tanh x=coshxsinhx
- 奇函数
- D o m ( c o s h x ) = R Dom(cosh\ x) = R Dom(cosh x)=R and I m ( c o s h x ) = ( 1 , 1 ) Im(cosh\ x) = (1, 1) Im(cosh x)=(1,1)
关系式
c o s h 2 x − s i n h 2 x = 1 s i n h ( x ± y ) = s i n h x c o s h y ± s i n h x c o s h y c o n s h ( x ± y ) = c o s h x c o s h y ± s i n h x s i n h y \begin{aligned} cosh^2 x - sinh^2 x & = 1\\ sinh(x \pm y) & = sinhx \ coshy \pm sinhx \ coshy \\ consh(x \pm y) & = coshx \ coshy \pm sinhx \ sinhy \\ \end{aligned} cosh2x−sinh2xsinh(x±y)consh(x±y)=1=sinhx coshy±sinhx coshy=coshx coshy±sinhx sinhy
反函数
s i n h − 1 x = l n ∣ x + x 2 + 1 ∣ c o s h − 1 x = l n ∣ x + x 2 − 1 ∣ t a n h − 1 x = 1 2 l n ∣ 1 + x 1 − x ∣ \begin{aligned} sinh^{-1}x & = ln \left| x + \sqrt {x^2 + 1} \right| \\ cosh^{-1}x & = ln \left| x + \sqrt {x^2 - 1} \right| \\ tanh^{-1}x & = \frac{1}{2} ln \left| \frac{1+x}{1-x} \right| \\ \end{aligned} sinh−1xcosh−1xtanh−1x=ln∣∣∣x+x2+1∣∣∣=ln∣∣∣x+x2−1∣∣∣=21ln∣∣∣∣1−x1+x∣∣∣∣
1.4 Differentiation 微分
- 莱布尼兹表示
d f d x \frac{df}{dx} dxdf - 格朗日表示
f ′ ( x ) f'(x) f′(x)
定义
f ′ ( x ) = lim δ x → 0 f ( x + δ x ) − f ( x ) δ x f'(x) = \lim_{\delta x \to 0} \frac{f(x + \delta x) - f(x)}{\delta x} f′(x)=δx→0limδxf(x+δx)−f(x)
上述定义式为前向微分
- 反向微分
f ′ ( x ) = lim δ x → 0 f ( x ) − f ( x − δ x ) δ x f'(x) = \lim_{\delta x \to 0} \frac{f(x) - f(x - \delta x)}{\delta x} f′(x)=δx→0limδxf(x)−f(x−δx) - 中心微分
f ′ ( x ) = lim δ x → 0 f ( x + δ x ) − f ( x − δ x ) 2 δ x f'(x) = \lim_{\delta x \to 0} \frac{f(x + \delta x) - f(x - \delta x)}{2 \delta x} f′(x)=δx→0lim2δxf(x+δx)−f(x−δx)
求微分的例子: f ( x ) = x 3 f(x) = x^3 f(x)=x3
f ′ ( x ) = lim δ x → 0 f ( x + δ x ) − f ( x ) δ x = lim δ x → 0 ( x + δ x ) 3 − x 3 δ x = lim δ x → 0 ( x 3 + 3 x 2 δ x + 3 x δ x 2 + δ x 3 ) − x 3 δ x = lim δ x → 0 ( 3 x 2 + 3 x δ x + δ x 2 ) = 3 x 2 \begin{aligned} f'(x) & = \lim_{\delta x \to 0} \frac{f(x + \delta x) - f(x)}{\delta x} \\ & = \lim_{\delta x \to 0} \frac{(x + \delta x)^3 - x^3}{\delta x} \\ & = \lim_{\delta x \to 0} \frac{(x^3 + 3 x^2 \delta x + 3x \delta x ^ 2 + \delta x ^ 3) - x^3}{\delta x} \\ & = \lim_{\delta x \to 0} (3x^2 + 3x\delta x + \delta x ^2) \\ & = 3x^2 \\ \end{aligned} f′(x)=δx→0limδxf(x+δx)−f(x)=δx→0limδx(x+δx)3−x3=δx→0limδx(x3+3x2δx+3xδx2+δx3)−x3=δx→0lim(3x2+3xδx+δx2)=3x2
常用微分
d d x x n = n x n − 1 d d x e x = e x d d x e a x = a e a x d d x l o g x = 1 x d d x s i n x = c o s x d d x c o s x = − s i n x d d x t a n x = s e c 2 x \frac{d}{dx} x^n = nx^{n-1} \\ \frac{d}{dx} e^x = e^x \\ \frac{d}{dx} e^{ax} = ae^{ax} \\ \frac{d}{dx} logx = \frac{1}{x} \\ \frac{d}{dx} sinx = cosx \\ \frac{d}{dx} cosx = -sinx \\ \frac{d}{dx} tanx = sec^2 x \\ dxdxn=nxn−1dxdex=exdxdeax=aeaxdxdlogx=x1dxdsinx=cosxdxdcosx=−sinxdxdtanx=sec2x
Linearity 线性:
线性的定义
- O p ( c × f ( x ) = c × O p ( f ( x ) ) Op(c \times f(x) = c \times Op(f(x)) Op(c×f(x)=c×Op(f(x))
- O p ( f ( x ) ± g ( x ) ) = O p ( f ( x ) ) ± O p ( g ( x ) ) Op(f(x) \pm g(x)) = Op(f(x)) \pm Op(g(x)) Op(f(x)±g(x))=Op(f(x))±Op(g(x))
两个函数加权和的微分,等于两个函数微分的加权和
y
=
λ
f
(
x
)
+
μ
g
(
x
)
,
λ
和
μ
是
常
量
d
y
d
x
=
d
d
x
(
λ
f
(
x
)
+
μ
g
(
x
)
)
=
λ
f
′
(
x
)
+
μ
g
′
(
x
)
y = \lambda f(x) + \mu g(x), \lambda和\mu是常量 \\ \frac{dy}{dx}=\frac{d}{dx} ( \lambda f(x) + \mu g(x))= \lambda f'(x) + \mu g'(x)
y=λf(x)+μg(x),λ和μ是常量dxdy=dxd(λf(x)+μg(x))=λf′(x)+μg′(x)
1.4.1 Product Rule 乘法法则
对两个函数的乘积求导数
y = f ( x ) × g ( x ) ⟹ d y d x = f ′ ( x ) × g ( x ) + f ( x ) × g ′ ( x ) y = f(x) \times g(x) \Longrightarrow \frac{dy}{dx} = f'(x) \times g(x) + f(x) \times g'(x) y=f(x)×g(x)⟹dxdy=f′(x)×g(x)+f(x)×g′(x)
1.4.2 Function of a Function Rule 函数的函数法则
求函数的函数的导数:chain rule 链式求导法则
y = f ( g ( x ) ) ⟹ d y d x = f ′ ( g ( x ) ) × g ′ ( x ) y = f(g(x)) \Longrightarrow \frac{dy}{dx} = f'(g(x)) \times g'(x) y=f(g(x))⟹dxdy=f′(g(x))×g′(x)
d y d x = d y d u × d u d x \frac{dy}{dx} = \frac{dy}{du} \times \frac{du}{dx} dxdy=dudy×dxdu
1.4.3 Quotient Rule 商数法则
可以用乘法法则推导
f ( x ) g ( x ) ⟹ d y d x = f ′ ( x ) × g ( x ) − f ( x ) × g ′ ( x ) ( ( g ( x ) ) 2 \frac{f(x)}{g(x)} \Longrightarrow \frac{dy}{dx} = \frac{f'(x) \times g(x) - f(x) \times g'(x)}{((g(x))^2} g(x)f(x)⟹dxdy=((g(x))2f′(x)×g(x)−f(x)×g′(x)
1.4.4 Implicit Differentiation 隐式微分
y = a x l n y = x l n a 1 y d y d x = l n a d y d x = a x l n a y = a^x \\ lny = xln \ a \\ \frac{1}{y} \frac{dy}{dx} = ln \ a \\ \frac{dy}{dx} = a^x ln \ a y=axlny=xln ay1dxdy=ln adxdy=axln a
1.4.5 Higher Derivatives 高阶导数
- n次多项式的 n+1 阶导数为0
- 不是所有函数都处处可导
1.4.6 Leibniz Rule 莱布尼兹法则
乘法法则的高阶表示(二项式)
D
n
(
u
v
)
=
∑
i
=
0
n
C
n
i
×
D
i
u
×
D
n
−
i
v
,
C
n
r
=
n
!
r
!
(
(
n
−
r
)
!
D^n(uv) = \sum_{i=0}^{n} C_n^i \times D^i u \times D^{n-i}v, \ C_n^r = \frac{n!}{r!((n-r)!}
Dn(uv)=i=0∑nCni×Diu×Dn−iv, Cnr=r!((n−r)!n!
1.4.7 Further Limits 高阶极限
L’Hospital’s rule
lim
x
→
a
f
(
x
g
(
x
)
≡
0
0
o
r
∞
∞
\lim_{x \to a} \frac{f(x}{g(x)} \equiv \frac{0}{0} \ or \ \frac{\infty}{\infty}
x→alimg(x)f(x≡00 or ∞∞
对极限的上下两部分同时求导,计算极限
lim x → a f ( x g ( x ) = lim x → a f ′ ( x g ′ ( x ) = … = lim x → a f ( r ) ( x g ( r ) ( x ) \lim_{x \to a} \frac{f(x}{g(x)} = \lim_{x \to a} \frac{f'(x}{g'(x)} = … = \lim_{x \to a} \frac{f^{(r)}(x}{g^{(r)}(x)} x→alimg(x)f(x=x→alimg′(x)f′(x=…=x→alimg(r)(x)f(r)(x
lim x → 0 s i n x x = lim x → 0 c o s x 1 = 1 \lim_{x \to 0} \frac{sinx}{x} = \lim_{x \to 0} \frac{cosx}{1} =1 \\ x→0limxsinx=x→0lim1cosx=1
1.5 Taylor Series 泰勒级数
多项式函数逼近原始函数,N次项的系数为N次导数
对于 f ( x ) = e x f(x)=e^x f(x)=ex, f ( r ) ( x ) = e x f^{(r)}(x) = e^x f(r)(x)=ex, f ( r ) ( 0 ) = 1 f^{(r)}(0) = 1 f(r)(0)=1
线性近似:
截距项为1,斜率为1,
e
x
≈
1
+
x
e^x \approx 1 + x
ex≈1+x
二次近似:
g
(
x
)
=
a
x
2
+
b
x
+
c
g
′
(
x
)
=
2
a
x
+
b
g
′
′
(
x
)
=
2
a
g(x)=ax^2+bx+c \\ g'(x)=2ax+b \\ g''(x)=2a
g(x)=ax2+bx+cg′(x)=2ax+bg′′(x)=2a
令
g
(
0
)
=
f
(
0
)
,
g
′
(
0
)
=
f
′
(
0
)
,
g
′
′
(
0
)
=
f
′
′
(
0
)
g(0)=f(0), g'(0)=f'(0), g''(0)=f''(0)
g(0)=f(0),g′(0)=f′(0),g′′(0)=f′′(0)
得到
c
=
1
,
b
=
1
,
a
=
1
2
c = 1, b = 1, a=\frac{1}{2}
c=1,b=1,a=21
因此
e
x
≈
1
+
x
+
1
2
x
2
e^x \approx 1 + x + \frac{1}{2} x^2
ex≈1+x+21x2
三次近似: e x ≈ 1 + x + 1 2 x 2 + 1 6 x 3 e^x \approx 1 + x + \frac{1}{2} x^2 + \frac{1}{6} x^3 ex≈1+x+21x2+61x3
无穷展开: e x = ∑ n = 0 ∞ x n n ! e^x = \sum_{n=0}^{\infty} \frac{x^n}{n!} ex=∑n=0∞n!xn
泰勒级数:
f
(
x
)
f(x)
f(x)在
x
0
x_0
x0的泰勒展开
f
(
x
)
=
∑
n
=
0
∞
1
n
!
f
(
n
)
(
x
0
)
(
x
−
x
0
)
n
f(x) = \sum_{n=0}^{\infty} \frac{1}{n!} f^{(n)}(x_0) \ (x-x_0)^n
f(x)=n=0∑∞n!1f(n)(x0) (x−x0)n
常用展开
e
x
=
∑
n
=
0
∞
x
n
n
!
l
o
g
(
1
+
x
)
=
∑
n
=
0
∞
(
−
1
)
n
−
1
x
n
n
!
e^x = \sum_{n=0}^{\infty} \frac{x^n}{n!} \\ log(1+x)= \sum_{n=0}^{\infty} (-1)^{n-1} \frac{x^n}{n!} \\
ex=n=0∑∞n!xnlog(1+x)=n=0∑∞(−1)n−1n!xn
泰勒定量
f ( x ) = ∑ k = 0 n − 1 1 k ! f ( k ) ( x 0 ) ( x − x 0 ) k + R n ( x ) 其 中 R n ( x ) = 1 n ! f ( n ) ( ξ ) ( x − x 0 ) n ξ 是 介 于 x 0 和 x 之 间 的 某 个 未 知 数 值 \begin{aligned} & f(x) = \sum_{k=0}^{n-1} \frac{1}{k!} f^{(k)}(x_0) \ (x-x_0)^k + R_n(x) \\ 其中 & R_n(x) = \frac{1}{n!} \ f^{(n)}(\xi) \ (x-x_0)^n \\ & \xi是介于x_0和x之间的某个未知数值 \end{aligned} 其中f(x)=k=0∑n−1k!1f(k)(x0) (x−x0)k+Rn(x)Rn(x)=n!1 f(n)(ξ) (x−x0)nξ是介于x0和x之间的某个未知数值
用泰勒展开式计算极限
lim
x
→
0
s
i
n
x
x
∼
lim
x
→
0
∑
i
=
0
∞
(
−
1
)
i
x
2
i
+
1
(
2
i
+
1
)
!
x
∼
lim
x
→
0
∑
i
=
0
∞
(
−
1
)
i
x
2
i
(
2
i
+
1
)
!
∼
lim
x
→
0
(
1
−
x
2
3
!
+
x
4
5
!
+
…
)
=
1
\begin{aligned} \lim_{x \to 0} \frac{sin \ x}{x} & \sim \lim_{x \to 0} \frac{\sum_{i=0}^{\infty} (-1)^{i} \frac{x^{2i+1}}{(2i+1)!}}{x} \\ & \sim \lim_{x \to 0} \sum_{i=0}^{\infty} (-1)^{i} \frac{x^{2i}}{(2i+1)!} \\ & \sim \lim_{x \to 0} (1 - \frac{x^2}{3!} + \frac{x^4}{5!} + …) \\ & = 1 \end{aligned}
x→0limxsin x∼x→0limx∑i=0∞(−1)i(2i+1)!x2i+1∼x→0limi=0∑∞(−1)i(2i+1)!x2i∼x→0lim(1−3!x2+5!x4+…)=1
1.5.1 The Binomial Expansion 二项式展开
二项式展开是 ( 1 + x ) n (1+x)^n (1+x)n的泰勒展开式
( 1 + x ) n = ∑ k = 0 n n ! k ! ( n − k ) ! x k ( 1 + a x ) n = ∑ k = 0 n n ! k ! ( n − k ) ! ( a x ) k ( p + a x ) n = ( p ( 1 + a p x ) ) n = p n ∑ k = 0 n n ! k ! ( n − k ) ! ( a p x ) k \begin{aligned} (1+x)^n & = \sum_{k=0}^n \frac{n!}{k! \ (n-k)!} x^k \\ (1+ax)^n & = \sum_{k=0}^n \frac{n!}{k! \ (n-k)!} (ax)^k \\ (p+ax)^n &= (p(1 + \frac{a}{p}x))^n\\ & = p^n \sum_{k=0}^n \frac{n!}{k! \ (n-k)!} (\frac{a}{p}x)^k\\ \end{aligned} (1+x)n(1+ax)n(p+ax)n=k=0∑nk! (n−k)!n!xk=k=0∑nk! (n−k)!n!(ax)k=(p(1+pax))n=pnk=0∑nk! (n−k)!n!(pax)k
Pascal三角形: ( 1 + x ) n (1+x)^n (1+x)n,不同的n的二项式系数组成的三角形
1.6 Integration 积分
1.6.1 The Indefinite Integral 不定积分
f ( x ) f(x) f(x)的不定积分 ∫ f ( x ) d x \int f(x)dx ∫f(x)dx
F ( x ) = ∫ f ( x ) d x d F ( x ) d x = f ( x ) F(x)=\int f(x)dx \\ \frac {dF(x)}{dx} = f(x) F(x)=∫f(x)dxdxdF(x)=f(x)
y = 2 x y=2x y=2x, d y d x = 2 \frac{dy}{dx}=2 dxdy=2, ∫ 2 d x = 2 x + C \int 2dx = 2x +C ∫2dx=2x+C, 注意常数项C对x的微分为0
不定积分的例子
∫
x
n
d
x
=
1
n
+
1
x
n
+
1
+
C
,
(
n
≠
−
1
)
∫
1
x
d
x
=
l
n
(
x
)
+
C
∫
e
a
x
d
x
=
1
a
e
a
x
+
C
∫
c
o
s
(
a
x
)
d
x
=
1
a
s
i
n
(
a
x
)
+
C
∫
s
i
n
(
a
x
)
d
x
=
−
1
a
c
o
s
(
a
x
)
+
C
\begin{aligned} \int x^n dx & = \frac{1}{n+1} x^{n+1} + C, \ (n \neq -1) \\ \int \frac{1}{x} dx & = ln(x) + C \\ \int e^{ax} dx & = \frac{1}{a} e^{ax} + C \\ \int cos(ax)dx & = \frac{1}{a}sin(ax) + C \\ \int sin(ax)dx & = -\frac{1}{a}cos(ax) + C \\ \end{aligned}
∫xndx∫x1dx∫eaxdx∫cos(ax)dx∫sin(ax)dx=n+11xn+1+C, (n=−1)=ln(x)+C=a1eax+C=a1sin(ax)+C=−a1cos(ax)+C
Linearity 线性:积分是线性的
∫
(
α
f
(
x
)
+
β
g
(
x
)
)
d
x
=
α
∫
f
(
x
)
d
x
+
β
∫
g
(
x
)
d
x
\int (\alpha f(x) + \beta g(x))dx = \alpha \int f(x)dx + \beta \int g(x)dx
∫(αf(x)+βg(x))dx=α∫f(x)dx+β∫g(x)dx
∫ ( A x 2 + B x 4 ) d x = A ∫ x 2 d x + B ∫ x 3 d x = A 3 x 3 B 4 x 4 + C ∫ ( 3 e x + 2 x ) d x = 3 ∫ e x d x + 2 ∫ 1 x d x = 3 e x + 2 l n x + C \int (A x^2 + B x^4)dx = A \int x^2 dx + B \int x^3 dx = \frac{A}{3} x^3 \frac{B}{4} x^4 + C \\ \int (3 e^x + \frac{2}{x})dx = 3 \int e^x dx + 2 \int \frac{1}{x} dx = 3e^x + 2lnx + C ∫(Ax2+Bx4)dx=A∫x2dx+B∫x3dx=3Ax34Bx4+C∫(3ex+x2)dx=3∫exdx+2∫x1dx=3ex+2lnx+C
1.6.2 The Definite Integral 定积分
f ( x ) f(x) f(x)的定积分 ∫ a b f ( x ) d x \int_a^b f(x)dx ∫abf(x)dx
例子:
∫
−
1
1
e
x
d
x
=
e
x
∣
−
1
1
=
e
−
1
e
\int _{-1}^{1} e^x dx = e^x | _{-1}^{1} = e - \frac{1}{e}
∫−11exdx=ex∣−11=e−e1
∫ a x f ( x ) d x \int_a^x f(x)dx ∫axf(x)dx 这种表示法,容易造成混淆,应使用哑变量
∫ a c f ( x ) d x = ∫ a b f ( x ) d x + ∫ b c f ( x ) d x \int_a^c f(x)dx = \int_a^b f(x)dx + \int_b^c f(x)dx ∫acf(x)dx=∫abf(x)dx+∫bcf(x)dx, if a < b < c a < b < c a<b<c
∫ a c f ( x ) d x = − ∫ c a f ( x ) d x \int_a^c f(x)dx = - \int_c^a f(x)dx ∫acf(x)dx=−∫caf(x)dx
1.6.3 Integration by Substitution 代换积分法/换元法
∫ g ( f ( x ) ) f ′ ( x ) d x \int g(f(x))f'(x)dx ∫g(f(x))f′(x)dx
反向使用链式法则:
令
z
=
f
(
x
)
z=f(x)
z=f(x)
则
d
z
=
f
′
(
x
)
d
x
可
得
∫
g
(
f
(
x
)
)
f
′
(
x
)
d
x
=
∫
g
(
z
)
d
z
dz=f'(x)dx 可得\int g(f(x))f'(x)dx = \int g(z)dz
dz=f′(x)dx可得∫g(f(x))f′(x)dx=∫g(z)dz
例子:
∫
1
2
e
x
2
2
x
d
x
\int_1^2 e^{x^2}2xdx
∫12ex22xdx,
令
z
=
x
2
z=x^2
z=x2,
可得
∫
1
2
e
x
2
2
x
d
x
=
∫
1
4
e
z
d
z
=
e
z
∣
1
4
=
e
4
−
e
1
\int_1^2 e^{x^2}2xdx = \int_1^4 e^zdz = e^z | _1^4 = e^4-e^1
∫12ex22xdx=∫14ezdz=ez∣14=e4−e1
重要例子:
∫
−
∞
∞
e
−
x
2
d
x
\int_{-\infty}^{\infty} e^{-x^2}dx
∫−∞∞e−x2dx
标准正态分布
Z
(
x
)
=
1
2
π
e
−
x
2
2
Z(x) = \frac{1}{\sqrt{2 \pi}}e^{-\frac{x^2}{2}}
Z(x)=2π1e−2x2
标准正态分布的CDF
∫
−
∞
∞
Z
(
x
)
d
x
=
1
2
π
∫
−
∞
∞
e
−
s
2
2
d
s
=
1
\int_{-\infty}^{\infty} Z(x)dx = \frac{1}{\sqrt{2 \pi}}\int_{-\infty}^{\infty} e^{-\frac{s^2}{2}}ds = 1
∫−∞∞Z(x)dx=2π1∫−∞∞e−2s2ds=1
令
x
=
s
2
x=\frac{s}{\sqrt{2}}
x=2s
可得
∫
−
∞
∞
e
−
x
2
d
x
=
π
\int_{-\infty}^{\infty} e^{-x^2}dx = \sqrt{\pi}
∫−∞∞e−x2dx=π
奇函数和偶函数的积分:
∫
−
a
a
f
(
x
)
d
x
=
∫
−
a
0
f
(
x
)
d
x
+
∫
0
a
f
(
x
)
d
x
=
−
∫
0
−
a
f
(
x
)
d
x
+
∫
0
a
f
(
x
)
d
x
=
∫
0
a
f
(
−
x
)
d
x
+
∫
0
a
f
(
x
)
d
x
=
{
2
∫
0
a
f
(
x
)
d
x
f(x)是偶函数
0
f(x)是奇函数
\begin{aligned} \int_{-a}^{a} f(x)dx & = \int_{-a}^{0} f(x)dx + \int_{0}^{a} f(x)dx \\ & = -\int_{0}^{-a} f(x)dx + \int_{0}^{a} f(x)dx \\ & = \int_{0}^{a} f(-x)dx + \int_{0}^{a} f(x)dx \\ & = \begin{cases} 2 \int_{0}^{a} f(x)dx & \text{f(x)是偶函数}\\ 0 & \text{f(x)是奇函数} \end{cases} \end{aligned}
∫−aaf(x)dx=∫−a0f(x)dx+∫0af(x)dx=−∫0−af(x)dx+∫0af(x)dx=∫0af(−x)dx+∫0af(x)dx={2∫0af(x)dx0f(x)是偶函数f(x)是奇函数
1.6.4 Integration by Parts 分部积分法
∫ u ′ v d x \int u'vdx ∫u′vdx
反向使用乘法法则
∫
u
′
v
d
x
=
u
v
−
∫
u
v
′
d
x
+
C
\int u'vdx = uv - \int uv'dx + C
∫u′vdx=uv−∫uv′dx+C
分部积分法的使用场景:v是多项式函数,u是指数函数
例子:
∫
x
e
x
d
x
\int xe^xdx
∫xexdx
令:
v
=
x
,
v
′
=
1
,
u
=
e
x
,
u
′
=
e
x
v=x, v'=1, u=e^x, u'=e^x
v=x,v′=1,u=ex,u′=ex
得到:
∫
x
e
x
d
x
=
u
v
−
∫
u
v
′
d
x
+
C
=
x
e
x
−
∫
e
x
×
1
d
x
=
e
x
(
x
−
1
)
+
C
\int xe^xdx = uv - \int uv'dx + C = xe^x - \int e^x \times 1dx = e^x(x-1) + C
∫xexdx=uv−∫uv′dx+C=xex−∫ex×1dx=ex(x−1)+C
弱国多项式部分的次数大于1,反复使用分部积分法直到变为0次
经典问题:
∫
e
x
s
i
n
x
d
x
\int e^x sinx dx
∫exsinxdx
令
v
=
e
x
,
u
′
=
s
i
n
x
,
v
′
=
e
x
,
u
=
−
c
o
s
x
v=e^x, u'=sinx, v'=e^x, u=-cosx
v=ex,u′=sinx,v′=ex,u=−cosx
得到:
∫
e
x
s
i
n
x
d
x
=
−
e
x
c
o
s
x
+
∫
e
x
c
o
s
x
d
x
\int e^x sinx dx = -e^xcosx + \int e^x cosx dx
∫exsinxdx=−excosx+∫excosxdx
令
v
=
e
x
,
u
′
=
c
o
s
x
,
v
′
=
e
x
,
u
=
s
i
n
x
v=e^x, u'=cosx, v'=e^x, u=sinx
v=ex,u′=cosx,v′=ex,u=sinx
得到:
∫
e
x
c
o
s
x
d
x
=
e
x
s
i
n
x
−
∫
e
x
s
i
n
x
d
x
\int e^x cosx dx = e^xsinx - \int e^x sinx dx
∫excosxdx=exsinx−∫exsinxdx
将上两个等式相加,消去
∫
e
x
c
o
s
x
d
x
\int e^x cosx dx
∫excosxdx项得到
∫
e
x
s
i
n
x
d
x
=
1
2
e
x
(
s
i
n
x
−
c
o
s
x
)
\int e^x sinx dx=\frac{1}{2}e^x(sinx-cosx)
∫exsinxdx=21ex(sinx−cosx)
∫
e
x
c
o
s
x
d
x
=
1
2
e
x
(
s
i
n
x
+
c
o
s
x
)
\int e^x cosx dx=\frac{1}{2}e^x(sinx+cosx)
∫excosxdx=21ex(sinx+cosx)
1.6.5 Reduction Formula 约化公式
∫ 0 ∞ e − t t n d t = I n \int_{0}^{\infty} e^{-t}t^ndt = I_n ∫0∞e−ttndt=In
用部分积分法逐次消去多项式项,得到
I
n
=
n
!
I
0
I_n = n!I_0
In=n!I0, 注意
e
−
t
t
n
∣
0
∞
=
0
e^{-t}t^n|_{0}^{\infty}=0
e−ttn∣0∞=0
I
0
=
∫
0
∞
e
−
t
d
t
=
1
I_0 = \int_{0}^{\infty} e^{-t}dt = 1
I0=∫0∞e−tdt=1,
I
n
=
n
!
I_n = n!
In=n!
I
n
I_n
In称为Gamma函数(
Γ
\Gamma
Γ函数)
1.6.6 Other Results 其他法则
∫
f
′
(
x
)
f
(
x
)
=
l
n
∣
f
(
x
)
∣
+
C
\int \frac{f'(x)}{f(x)} = ln|f(x)| +C
∫f(x)f′(x)=ln∣f(x)∣+C
∫
1
a
+
b
x
=
1
b
l
n
∣
a
+
b
x
∣
+
C
\int \frac{1}{a+bx} = \frac{1}{b}ln|a+bx| + C
∫a+bx1=b1ln∣a+bx∣+C
Partial Fractions:部分分式分解
h
(
x
)
=
f
(
x
)
g
(
x
)
=
∑
n
=
0
N
a
n
x
n
∑
n
=
0
M
b
n
x
n
h(x) = \frac{f(x)}{g(x)} = \frac{\sum_{n=0}^{N} a_n x^n}{\sum_{n=0}^{M} b_n x^n}
h(x)=g(x)f(x)=∑n=0Mbnxn∑n=0Nanxn
如果N<M,则h(x)称为部分分式分解
c
(
x
+
a
)
(
x
+
b
)
≡
A
x
+
a
+
B
x
+
b
c
=
A
(
x
+
b
)
+
B
(
x
+
a
)
\frac{c}{(x+a)(x+b)} \equiv \frac{A}{x+a} + \frac{B}{x+b} \\ c = A(x+b) + B(x+a)
(x+a)(x+b)c≡x+aA+x+bBc=A(x+b)+B(x+a)
求解出A和B,得到部分分式分解
-
重复的因式:
c ( x + a ) 2 ( x + b ) 3 = A x + a + B ( x + a ) 2 + C x + b + D ( x + b ) 2 + E ( x + b ) 3 \frac{c}{(x+a)^2(x+b)^3} = \frac{A}{x+a} + \frac{B}{(x+a)^2} + \frac{C}{x+b} + \frac{D}{(x+b)^2} + \frac{E}{(x+b)^3} (x+a)2(x+b)3c=x+aA+(x+a)2B+x+bC+(x+b)2D+(x+b)3E -
未分解的高次项
2 x + 1 ( x 2 + 3 x + 2 ) ( x − 1 ) = A x + B x 2 + 3 x + 2 + C x − 1 \frac{2x+1}{(x^2+3x+2)(x-1)} = \frac{Ax+B}{x^2+3x+2} + \frac{C}{x-1} (x2+3x+2)(x−1)2x+1=x2+3x+2Ax+B+x−1C
1.7 Complex Numbers 复数
复数的定义
z
=
x
+
i
y
z = x + iy
z=x+iy where
x
,
y
∈
R
x, y \in R
x,y∈R and
i
=
−
1
i = \sqrt{-1}
i=−1
x
x
x称为实部real part,
y
y
y称为虚部imaginary part
极坐标表示形式
z
=
r
(
c
o
s
θ
+
i
s
i
n
θ
)
z = r(cos\theta + i \ sin\theta)
z=r(cosθ+i sinθ)
x
=
r
c
o
s
θ
,
y
=
r
s
i
n
θ
)
,
θ
=
a
r
c
t
a
n
y
x
x = r \ cos\theta, \ y =r \ sin\theta), \ \theta = arctan \frac{y}{x}
x=r cosθ, y=r sinθ), θ=arctanxy
共轭conjugate
z
=
x
+
i
y
z=x+iy
z=x+iy和
z
=
x
−
i
y
z=x-iy
z=x−iy互为共轭复数
1.7.1 Arithmetic 算术运算
- 加减法: z 1 ± z 2 = ( x 1 ± x 2 ) + i ( y 1 ± y 2 ) ) z_1 \pm z_2 = (x_1 \pm x_2) + i(y_1 \pm y_2)) z1±z2=(x1±x2)+i(y1±y2))
- 乘法: z 1 × z 2 = ( x 1 x 2 − y 1 y 2 ) + i ( x 1 y 2 + x 2 y 1 ) z_1 \times z_2 = (x_1 x_2 - y_1 y_2) + i(x_1 y_2 + x_2 y_1) z1×z2=(x1x2−y1y2)+i(x1y2+x2y1)
- 除法:
z
1
z
2
=
1
x
2
2
+
y
2
2
(
(
x
1
x
2
+
y
1
y
2
)
+
i
(
x
1
y
2
−
x
2
y
1
)
)
\frac{z_1}{z_2} = \frac{1}{x_2 ^ 2 + y_2 ^ 2} ((x_1 x_2 + y_1 y_2) + i(x_1 y_2 - x_2 y_1))
z2z1=x22+y221((x1x2+y1y2)+i(x1y2−x2y1))
除法相当于上下都乘以 z 2 z_2 z2的共轭复数 x 2 − i y 2 x_2 - i \ y_2 x2−i y2
1.7.2 Complex Conjugate Identities 共轭复数的性质
- 共轭的共轭 ( z ˉ ) ‾ = z \overline {(\bar z)} = z (zˉ)=z
- 加法的共轭 ( z 1 + z 2 ) ‾ = z 1 ‾ + z 2 ‾ \overline {(z_1 + z_2)} = \overline {z_1} + \overline {z_2} (z1+z2)=z1+z2
- 乘积的共轭 ( z 1 × z 2 ) ‾ = z 1 ‾ × z 2 ‾ \overline {(z_1 \times z_2)} = \overline {z_1} \times \overline {z_2} (z1×z2)=z1×z2
- 共轭相加 z + z ˉ = 2 x = 2 R e z R e z = z + z ˉ 2 z + \bar z = 2x = 2 Re \ z \\ Re \ z = \frac{z + \bar z}{2} z+zˉ=2x=2Re zRe z=2z+zˉ
- 共轭相减 z − z ˉ = 2 i y = 2 i I m z I m z = z − z ˉ 2 i z - \bar z = 2iy = 2i Im \ z \\ Im \ z = \frac{z - \bar z}{2i} z−zˉ=2iy=2iIm zIm z=2iz−zˉ
- 共轭相乘 z × z ˉ = ( x + i y ) ( x − i y ) = ∣ z ∣ 2 z \times \bar z = (x + iy)(x - iy) = |z|^2 z×zˉ=(x+iy)(x−iy)=∣z∣2
1.7.3 Polar Form 极坐标形式
z = r ( c o s θ + i s i n θ ) = r e i θ e i θ = c o s θ + i s i n θ z = r(cos \theta + i \ sin\theta) = re^{i \theta} \\ e^{i \theta} = cos \theta + i \ sin\theta z=r(cosθ+i sinθ)=reiθeiθ=cosθ+i sinθ
极坐标形式的乘除法非常简便
Euler’s Formula 欧拉公式
可以通过泰勒级数证明 e i θ = c o s θ + i s i n θ e^{i \theta} = cos \theta + i \ sin\theta eiθ=cosθ+i sinθ,关键点在于 i 2 = − 1 i^2=-1 i2=−1
泰勒展开式
e
x
=
∑
n
=
0
∞
x
n
n
!
s
i
n
x
=
∑
n
=
0
∞
(
−
1
)
n
x
(
2
n
+
1
)
(
2
n
+
1
)
!
c
o
s
x
=
∑
n
=
0
∞
(
−
1
)
n
x
(
2
n
)
(
2
n
)
!
\begin{aligned} e^x & = \sum_{n=0}^{\infty} \frac{x^n}{n!} \\ sin \ x & = \sum_{n=0}^{\infty} (-1)^n \frac{x^{(2n+1)}}{(2n+1)!} \\ cos \ x & = \sum_{n=0}^{\infty} (-1)^n \frac{x^{(2n)}}{(2n)!} \\ \end{aligned}
exsin xcos x=n=0∑∞n!xn=n=0∑∞(−1)n(2n+1)!x(2n+1)=n=0∑∞(−1)n(2n)!x(2n)
欧拉公式的证明
e
i
θ
=
∑
n
=
0
∞
(
i
θ
)
n
n
!
=
∑
n
=
0
∞
(
i
θ
)
(
2
n
)
(
2
n
)
!
+
∑
n
=
0
∞
(
i
θ
)
(
2
n
+
1
)
(
2
n
+
1
)
!
=
∑
n
=
0
∞
i
(
2
n
)
θ
(
2
n
)
(
2
n
)
!
+
∑
n
=
0
∞
i
×
i
(
2
n
)
θ
(
2
n
+
1
)
(
2
n
+
1
)
!
=
∑
n
=
0
∞
(
−
1
)
n
x
(
2
n
)
(
2
n
)
!
+
i
∑
n
=
0
∞
(
−
1
)
n
x
(
2
n
+
1
)
(
2
n
+
1
)
!
=
c
o
s
θ
+
i
s
i
n
θ
\begin{aligned} e^{i \theta} & = \sum_{n=0}^{\infty} \frac{(i \theta)^n}{n!} \\ & = \sum_{n=0}^{\infty} \frac{(i \theta)^{(2n)}}{(2n)!} + \sum_{n=0}^{\infty} \frac{(i \theta)^{(2n+1)}}{(2n+1)!} \\ & = \sum_{n=0}^{\infty} \frac{i^{(2n)} \theta^{(2n)}}{(2n)!} + \sum_{n=0}^{\infty} \frac{i \times i^{(2n)} \theta^{(2n+1)}}{(2n+1)!} \\ &= \sum_{n=0}^{\infty} (-1)^n \frac{x^{(2n)}}{(2n)!} + i \sum_{n=0}^{\infty} (-1)^n \frac{x^{(2n+1)}}{(2n+1)!} \\ &= cos \theta + i sin \theta \\ \end{aligned}
eiθ=n=0∑∞n!(iθ)n=n=0∑∞(2n)!(iθ)(2n)+n=0∑∞(2n+1)!(iθ)(2n+1)=n=0∑∞(2n)!i(2n)θ(2n)+n=0∑∞(2n+1)!i×i(2n)θ(2n+1)=n=0∑∞(−1)n(2n)!x(2n)+in=0∑∞(−1)n(2n+1)!x(2n+1)=cosθ+isinθ
用欧拉公式计算
∫
e
x
s
i
n
x
d
x
\int e^x sinx dx
∫exsinxdx
∫
e
x
s
i
n
x
d
x
=
∫
e
x
I
m
e
i
x
d
x
=
∫
I
m
e
(
1
+
i
)
x
d
x
=
I
m
1
1
+
i
e
(
1
+
i
)
x
d
x
=
e
x
I
m
1
−
i
2
e
i
x
d
x
=
1
2
e
x
I
m
(
1
−
i
)
(
c
o
s
x
+
i
s
i
n
x
)
=
1
2
e
x
(
s
i
n
x
−
c
o
s
x
)
\begin{aligned} \int e^x sinx dx & = \int e^x Im e^{ix} dx \\ & = \int Im e^{(1 + i)x} dx \\ & = Im \frac{1}{1+i}e^{(1 + i)x} dx \\ & = e^x Im \frac{1-i}{2}e^{ix} dx \\ & = \frac{1}{2} e^x Im (1 - i)(cosx+isinx) \\ & = \frac{1}{2} e^x (sinx - cosx) \end{aligned}
∫exsinxdx=∫exImeixdx=∫Ime(1+i)xdx=Im1+i1e(1+i)xdx=exIm21−ieixdx=21exIm(1−i)(cosx+isinx)=21ex(sinx−cosx)
同样的方法可以得到 ∫ e x s i n x d x = 1 2 e x ( s i n x + c o s x ) \int e^x sinx dx = \frac{1}{2} e^x (sinx + cosx) ∫exsinxdx=21ex(sinx+cosx)
1.8 Functions of Several Variables: Multivariate Calculus 多变量函数:多变量微积分
偏微分 partial derivative
对
f
(
x
,
y
)
f(x, y)
f(x,y)定义偏微分
∂
f
∂
x
=
l
i
m
δ
x
→
0
f
(
x
+
δ
x
,
y
)
−
f
(
x
,
y
)
δ
x
\frac {\partial f}{\partial x} = lim_{\delta x \to 0} \frac{f(x + \delta x, y) - f(x, y)}{\delta x}
∂x∂f=limδx→0δxf(x+δx,y)−f(x,y)
其中y保持不变(看作常量)
偏微分也记作
f
x
,
f
y
f_x, \ f_y
fx, fy
高阶偏微分
∂
2
f
∂
x
2
=
f
x
x
=
∂
∂
x
(
∂
f
∂
x
)
∂
2
f
∂
y
2
=
f
y
y
=
∂
∂
y
(
∂
f
∂
y
)
∂
2
f
∂
x
∂
y
=
f
x
y
=
∂
∂
y
(
∂
f
∂
x
)
∂
2
f
∂
y
∂
x
=
f
y
x
=
∂
∂
x
(
∂
f
∂
y
)
\frac {\partial ^2 f}{\partial x^2} = f_{xx} = \frac {\partial}{\partial x} (\frac {\partial f}{\partial x}) \\ \frac {\partial ^2 f}{\partial y^2} = f_{yy} = \frac {\partial}{\partial y} (\frac {\partial f}{\partial y}) \\ \frac {\partial ^2 f}{\partial x \partial y} = f_{xy} = \frac {\partial}{\partial y} (\frac {\partial f}{\partial x}) \\ \frac {\partial ^2 f}{ \partial y \partial x} = f_{yx} = \frac {\partial}{\partial x} (\frac {\partial f}{\partial y}) \\
∂x2∂2f=fxx=∂x∂(∂x∂f)∂y2∂2f=fyy=∂y∂(∂y∂f)∂x∂y∂2f=fxy=∂y∂(∂x∂f)∂y∂x∂2f=fyx=∂x∂(∂y∂f)
1.8.1 The Chain Rule I 链式法则I
单变量:
f
(
u
)
,
u
=
g
(
x
)
d
f
d
x
=
d
f
d
u
×
d
u
d
x
f(u), \ u=g(x) \\ \frac{df}{dx} = \frac{df}{du} \times \frac{du}{dx}
f(u), u=g(x)dxdf=dudf×dxdu
多变量: 多个变量都是某个最终变量的函数
f
(
x
,
y
)
,
x
=
x
(
s
)
,
y
=
y
(
s
)
d
f
d
s
=
∂
f
∂
x
×
d
x
d
s
+
∂
f
∂
y
×
d
y
d
s
f(x, y), \ x=x(s), \ y=y(s) \\ \frac{df}{ds} = \frac{\partial f}{\partial x} \times \frac{dx}{ds} + \frac{\partial f}{\partial y} \times \frac{dy}{ds}
f(x,y), x=x(s), y=y(s)dsdf=∂x∂f×dsdx+∂y∂f×dsdy
1.8.2 The Chain Rule II 链式法则II
多变量: 多个变量都是某一组最终变量的函数
f
(
x
,
y
)
,
x
=
x
(
u
,
v
)
,
y
=
y
(
u
,
v
)
d
f
d
u
=
∂
f
∂
x
×
∂
x
∂
u
+
∂
f
∂
y
×
∂
y
∂
u
f(x, y), \ x=x(u,v), \ y=y(u,v) \\ \frac{df}{du} = \frac{\partial f}{\partial x} \times \frac{\partial x}{\partial u} + \frac{\partial f}{\partial y} \times \frac{\partial y}{\partial u}
f(x,y), x=x(u,v), y=y(u,v)dudf=∂x∂f×∂u∂x+∂y∂f×∂u∂y
1.8.3 Taylor for two Variables 两变量泰勒展开
f ( x , t ) f(x,t) f(x,t)在 x = x 0 , t = t 0 x=x_0, \ t=t_0 x=x0, t=t0处展开
f ( x ) = f ( x 0 , t 0 ) + f x ( x 0 , t 0 ) ( x − x 0 ) + f t ( x 0 , t 0 ) ( t − t 0 ) + 1 2 { f x x ( x 0 , t 0 ) ( x − x 0 ) 2 + 2 f x t ( x 0 , t 0 ) ( x − x 0 ) ( t − t 0 ) + f t t ( x 0 , t 0 ) ( t − t 0 ) 2 } + … \begin{aligned} f(x) = & f(x_0, t_0) \\ & + f_x(x_0, t_0)(x - x_0) + f_t(x_0, t_0)(t - t_0) \\ & + \frac{1}{2} \left\{ \begin{aligned} & f_{xx}(x_0, t_0)(x - x_0)^2 \\ & + 2f_{xt}(x_0, t_0)(x - x_0)(t - t_0) \\ & + f_{tt}(x_0, t_0)(t - t_0)^2 \\ \end{aligned} \right\} \\ & + \dots \end{aligned} f(x)=f(x0,t0)+fx(x0,t0)(x−x0)+ft(x0,t0)(t−t0)+21⎩⎪⎨⎪⎧fxx(x0,t0)(x−x0)2+2fxt(x0,t0)(x−x0)(t−t0)+ftt(x0,t0)(t−t0)2⎭⎪⎬⎪⎫+…
2 Linear Algebra 线性代数
2.1 Properties of Vectors 向量的性质
n维空间
R
n
R_n
Rn
n维向量
v
⃗
=
[
v
1
v
2
⋮
v
n
]
∈
R
n
\vec \boldsymbol v = \left[ \begin{matrix} v_1 \\ v_2 \\ \vdots \\ v_n \\ \end{matrix} \right] \in R_n
v=⎣⎢⎢⎢⎡v1v2⋮vn⎦⎥⎥⎥⎤∈Rn
v
i
,
(
i
=
1
,
2
,
⋯
,
n
)
v_i, \ (i = 1, 2, \cdots, n)
vi, (i=1,2,⋯,n)是向量
v
v
v的第i个维度
2.1.1 Vector Arithmetic 向量算术
相同维度的向量和
u
⃗
+
v
⃗
=
[
u
1
+
v
1
u
2
+
v
2
⋮
u
n
+
v
n
]
\vec \boldsymbol u + \vec \boldsymbol v = \left[ \begin{matrix} u_1 + v_1 \\ u_2 + v_2 \\ \vdots \\ u_n + v_n \\ \end{matrix} \right]
u+v=⎣⎢⎢⎢⎡u1+v1u2+v2⋮un+vn⎦⎥⎥⎥⎤
向量和标量的乘法
k v ⃗ = [ k v 1 k v 2 ⋮ k v n ] k \vec \boldsymbol v = \left[ \begin{matrix} kv_1 \\ kv_2 \\ \vdots \\ kv_n \\ \end{matrix} \right] kv=⎣⎢⎢⎢⎡kv1kv2⋮kvn⎦⎥⎥⎥⎤
向量点积
u ⃗ . v ⃗ = ∑ i = 1 n u i v i \vec \boldsymbol u . \vec \boldsymbol v = \sum_{i=1}^{n} u_i v_i u.v=i=1∑nuivi
u ⃗ . v ⃗ = v ⃗ . u ⃗ \vec \boldsymbol u . \vec \boldsymbol v = \vec \boldsymbol v . \vec \boldsymbol u u.v=v.u
2.1.2 Concept of Length in R n R_n Rn n维空间中的向量长度
向量长度
∣
u
⃗
∣
=
∑
i
=
1
n
u
i
2
u
⃗
=
∣
u
⃗
∣
u
⃗
^
u
⃗
^
=
u
⃗
∣
u
⃗
∣
∣
u
⃗
^
∣
=
1
|\vec \boldsymbol u | = \sum_{i=1}^{n} u_i^2 \\ \vec \boldsymbol u = |\vec \boldsymbol u | \hat{\vec \boldsymbol u} \\ \hat{\vec \boldsymbol u}= \frac{\vec \boldsymbol u }{|\vec \boldsymbol u | } \\ |\hat{\vec \boldsymbol u}| = 1
∣u∣=i=1∑nui2u=∣u∣u^u^=∣u∣u∣u^∣=1
u ⃗ ^ \hat{\vec \boldsymbol u} u^是单位向量,反映向量的方向
向量距离
∣
u
⃗
−
v
⃗
∣
=
∑
i
=
1
n
(
u
i
−
v
i
)
2
|\vec \boldsymbol u - \vec \boldsymbol v| = \sum_{i=1}^{n} (u_i - v_i)^2
∣u−v∣=i=1∑n(ui−vi)2
向量 u ⃗ \vec \boldsymbol u u长度也叫L2 norm/欧几里得norm: ∥ u ⃗ ∥ = ∑ i = 1 n u i 2 \left \| \vec \boldsymbol u \right \| = \sum_{i=1}^{n} u_i^2 ∥u∥=∑i=1nui2
L2距离/欧几里得距离: ∥ u ⃗ − v ⃗ ∥ = ∑ i = 1 n ( u i − v i ) 2 \left \| \vec \boldsymbol u - \vec \boldsymbol v \right \| = \sum_{i=1}^{n} (u_i - v_i)^2 ∥u−v∥=∑i=1n(ui−vi)2
余弦夹角
u ⃗ . v ⃗ = ∣ u ⃗ ∣ ∣ v ⃗ ∣ c o s θ → c o s θ = ∑ i = 1 n u i v i ∣ u ⃗ ∣ ∣ v ⃗ ∣ \vec \boldsymbol u . \vec \boldsymbol v = |\vec \boldsymbol u| |\vec \boldsymbol v| cos \theta \\ \rightarrow cos \theta = \frac{\sum_{i=1}^{n} u_i v_i}{ |\vec \boldsymbol u| |\vec \boldsymbol v|} \\ u.v=∣u∣∣v∣cosθ→cosθ=∣u∣∣v∣∑i=1nuivi
余弦夹角可以理解为两个向量的归一化距离
向量正交:
- 两个向量的点积为0, u ⃗ . v ⃗ = 0 \vec \boldsymbol u . \vec \boldsymbol v=0 u.v=0
- 向量夹角为90°
2.2 Matrices 矩阵
A
m
×
n
\boldsymbol A_{m \times n}
Am×n 表示m行n列的矩阵
A
=
[
a
11
a
12
⋯
a
1
n
a
21
a
22
⋯
a
2
n
⋮
⋮
⋱
⋮
a
m
1
a
m
2
⋯
a
m
n
]
A
=
(
a
i
j
)
f
o
r
i
=
1
,
2
,
⋯
,
m
;
j
=
1
,
2
,
⋯
,
n
\boldsymbol A = \left[ \begin{matrix} a_{11} & a_{12} & \cdots & a_{1n} \\ a_{21} & a_{22} & \cdots & a_{2n} \\ \vdots & \vdots & \ddots & \vdots \\ a_{m1} & a_{m2} & \cdots & a_{mn} \\ \end{matrix} \right] \\ \boldsymbol A = (a_{ij}) \ for \ i = 1,2, \cdots, m; \ j = 1, 2, \cdots, n
A=⎣⎢⎢⎢⎡a11a21⋮am1a12a22⋮am2⋯⋯⋱⋯a1na2n⋮amn⎦⎥⎥⎥⎤A=(aij) for i=1,2,⋯,m; j=1,2,⋯,n
2.2.1 Matrix Arithmetic 矩阵算术
矩阵加法 : 对应元素相加
交换律 A + B = B + A \boldsymbol A + \boldsymbol B = \boldsymbol B + \boldsymbol A A+B=B+A
2.2.2 Matrix Multiplication 矩阵乘法
矩阵乘法:
C
p
m
=
A
p
n
B
n
m
\boldsymbol C_{pm} = \boldsymbol A_{pn} \boldsymbol B_{nm}
Cpm=ApnBnm
A的列数和B的行数应该相同
C i j = A i ⃗ . B j ⃗ = ∑ k = 1 N A i k B k j \boldsymbol C_{ij} = \vec{\boldsymbol A_i} . \vec{\boldsymbol B_j} = \sum_{k=1}^N \boldsymbol A_{ik} \boldsymbol B_{kj} Cij=Ai.Bj=k=1∑NAikBkj
2.2.3 Transpose 转置
矩阵的行列交换
( A + B ) T = A T + B T ( A T ) T = A ( k A ) T = k A T ( A B ) T = B T A T {(\boldsymbol A+\boldsymbol B)}^T = {\boldsymbol A}^T + {\boldsymbol B}^T \\ {({\boldsymbol A}^T)}^T = \boldsymbol A \\ {(k \boldsymbol A)}^T = k {\boldsymbol A}^T \\ {(\boldsymbol A \boldsymbol B)}^T = {\boldsymbol B}^T {\boldsymbol A}^T \\ (A+B)T=AT+BT(AT)T=A(kA)T=kAT(AB)T=BTAT
skew symmetric matrix 斜对称矩阵 : A + A T = 0 ; a i i = 0 , a i j = − a j i \boldsymbol A + {\boldsymbol A}^T = 0; a_{ii}=0, a_{ij} = -a_{ji} A+AT=0;aii=0,aij=−aji
2.2.4 Matrix Representation of Linear Equations 线性方程组的矩阵表达形式
线性方程组
{
a
x
+
b
y
=
p
c
x
+
d
y
=
q
\left\{ \begin{aligned} ax + by = p \\ cx + dy = q \\ \end{aligned} \right.
{ax+by=pcx+dy=q
当 a d − b c ≠ 0 ad - bc \neq 0 ad−bc=0时,方程组有唯一解
{ x = d p − b q a d − b c y = a q − c p a d − b c \left\{ \begin{aligned} x = \frac{dp - bq}{ad - bc} \\ y = \frac{aq - cp}{ad - bc} \\ \end{aligned} \right. ⎩⎪⎪⎨⎪⎪⎧x=ad−bcdp−bqy=ad−bcaq−cp
当 a d − b c = 0 ad - bc = 0 ad−bc=0时,可能无解,也可能有无穷多解
矩阵表示
[
a
b
c
d
]
×
[
x
y
]
=
[
p
q
]
A
x
⃗
=
p
⃗
\left[ \begin{matrix} a & b \\ c & d \\ \end{matrix} \right] \times \left[ \begin{matrix} x \\ y \\ \end{matrix} \right] = \left[ \begin{matrix} p \\ q \\ \end{matrix} \right] \\ \boldsymbol A \vec \boldsymbol x = \vec \boldsymbol p
[acbd]×[xy]=[pq]Ax=p
行列式 determinant
∣
a
b
c
d
∣
=
a
d
−
b
c
\left| \begin{matrix} a & b \\ c & d \\ \end{matrix} \right| = ad - bc
∣∣∣∣acbd∣∣∣∣=ad−bc
方阵行列式不为0,表示矩阵可逆
单位阵 identity matrix
∣
1
0
0
⋯
0
0
1
0
⋯
0
0
0
1
⋯
0
⋮
⋮
⋮
⋱
⋮
0
0
0
⋯
1
∣
=
a
d
−
b
c
\left| \begin{matrix} 1 & 0 & 0 & \cdots & 0 \\ 0 & 1 & 0 & \cdots & 0 \\ 0 & 0 & 1 & \cdots & 0 \\ \vdots & \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & 0 & \cdots & 1 \\ \end{matrix} \right| = ad - bc
∣∣∣∣∣∣∣∣∣∣∣100⋮0010⋮0001⋮0⋯⋯⋯⋱⋯000⋮1∣∣∣∣∣∣∣∣∣∣∣=ad−bc
I A = A I = A \boldsymbol I \boldsymbol A = \boldsymbol A \boldsymbol I = \boldsymbol A IA=AI=A
三对角矩阵(Tridiagonal Matrices)
从右到左三条对角线分别是:super、main、sub
2.3 Using Matrix Notation For Solving Linear Systems 矩阵术语
线性方程组
{
a
x
+
b
y
+
c
z
=
p
d
x
+
e
y
+
f
z
=
q
g
x
+
h
y
+
i
z
=
r
\left\{ \begin{aligned} ax + by + cz = p \\ dx + ey + fz = q \\ gx + hy + iz = r \\ \end{aligned} \right.
⎩⎪⎨⎪⎧ax+by+cz=pdx+ey+fz=qgx+hy+iz=r
[ a b c d e f g h i ] × [ x y z ] = [ p q r ] A x ⃗ = p ⃗ \left[ \begin{matrix} a & b & c \\ d & e & f \\ g & h & i \\ \end{matrix} \right] \times \left[ \begin{matrix} x \\ y \\ z \\ \end{matrix} \right] = \left[ \begin{matrix} p \\ q \\ r \\ \end{matrix} \right] \\ \boldsymbol A \vec \boldsymbol x = \vec \boldsymbol p ⎣⎡adgbehcfi⎦⎤×⎣⎡xyz⎦⎤=⎣⎡pqr⎦⎤Ax=p
A \boldsymbol A A是系数矩阵
增广矩阵 augmented matrix
[
a
b
c
d
e
f
g
h
i
p
q
r
]
\left [ \begin{array}{c|c} \begin{matrix} a & b & c \\ d & e & f \\ g & h & i \\ \end{matrix} & \begin{matrix} p \\ q \\ r \\ \end{matrix} \end{array} \right ]
⎣⎡adgbehcfipqr⎦⎤
行阶梯形式 echelon form,对增广矩阵化简,对角线以下部分全部是0
矩阵行操作(对方程组的解没有影响)
- [ E R 1 ] [ER_1] [ER1]: 行交换
- [ E R 2 ] [ER_2] [ER2]: 行乘以系数k
- [ E R 3 ] [ER_3] [ER3]: 行加上其他行乘以系数k
通过行操作将增广矩阵变换为行阶梯形式
2.4 Matrix Inverse 逆变换
A A − 1 = A − 1 A = I \boldsymbol A \boldsymbol A^ {-1} = \boldsymbol A^ {-1} \boldsymbol A = I AA−1=A−1A=I
A x ⃗ = p ⃗ → x = A − 1 p ⃗ \boldsymbol A \vec \boldsymbol x = \vec \boldsymbol p \rightarrow \boldsymbol x = \boldsymbol A^ {-1} \vec \boldsymbol p Ax=p→x=A−1p
( a b c d ) − 1 = 1 a d − b c ( d − b − c a ) , a d − b c ≠ 0 \left( \begin{matrix} a & b \\ c & d \\ \end{matrix} \right) ^ {-1} = \frac{1}{ad-bc} \left( \begin{matrix} d & -b \\ -c & a \\ \end{matrix} \right), \ ad-bc \neq 0 (acbd)−1=ad−bc1(d−c−ba), ad−bc=0
用行列式计算逆矩阵
A
−
1
=
1
∣
A
∣
a
d
j
A
a
d
j
A
=
[
(
−
1
)
i
+
j
∣
M
i
j
∣
]
T
\boldsymbol A^ {-1} = \frac{1}{|\boldsymbol A|} adj\ \boldsymbol A \\ adj\ \boldsymbol A = [ (-1)^{i+j} |\boldsymbol M_{ij}|]^T
A−1=∣A∣1adj Aadj A=[(−1)i+j∣Mij∣]T
(
−
1
)
i
+
j
M
i
j
(-1)^{i+j} \boldsymbol M_{ij}
(−1)i+jMij 是行列式的余子式cofactor,从矩阵A去掉i行j列剩下的部分组成的矩阵,
行列式计算规则:
- [ E R 1 ] : R i ↔ R j ⇒ ∣ A ∣ = − ∣ A ∣ [ER_1]: R_i \leftrightarrow R_j \Rightarrow |\boldsymbol A| = -|\boldsymbol A| [ER1]:Ri↔Rj⇒∣A∣=−∣A∣
- [ E R 2 ] : R i ↔ k R i ⇒ ∣ A ∣ = k ∣ A ∣ [ER_2]: R_i \leftrightarrow kR_i \Rightarrow |\boldsymbol A| = k|\boldsymbol A| [ER2]:Ri↔kRi⇒∣A∣=k∣A∣
2.5 Orthogonal Matrices 正交矩阵
A A T = A T A = I A − 1 = A T \boldsymbol A \boldsymbol A^T = \boldsymbol A^T \boldsymbol A = I \\ \boldsymbol A^ {-1} = \boldsymbol A^T AAT=ATA=IA−1=AT
2.6 Eigenvalues and Eigenvectors 特征值和特征向量
A v ⃗ = λ v ⃗ ( A − λ I ) v ⃗ = 0 d e t ( A − λ I ) = 0 \boldsymbol A \vec \boldsymbol v = \lambda \vec \boldsymbol v \\ (\boldsymbol A - \lambda \boldsymbol I) \vec \boldsymbol v = 0 \\ det(\boldsymbol A - \lambda \boldsymbol I) = 0 \\ Av=λv(A−λI)v=0det(A−λI)=0
2.6.1 Criteria for invertibility 可逆
- 行列式不为0
- 所有特征值都非零
- strictly diagonally dominant 严格对角占优(充分不必要)
3 Differential Equations 微分方程
3.1 Introduction 简介
两种类型
-
Ordinary Differential Equation (O.D.E) 常微分方程
F ( x , y , y ′ , y ′ ′ ′ , ⋯ , y ( n ) ) = 0 F(x, y, y',y''', \cdots, y^{(n)}) = 0 F(x,y,y′,y′′′,⋯,y(n))=0
y 4 ≠ y ( 4 ) y^4 \neq y^{(4)} y4=y(4) -
Partial Differential Equation (PDE) 偏微分方程
∂ u ∂ t + ∂ 2 u ∂ x ∂ y + ∂ u ∂ z + u = 0 \frac{\partial u}{\partial t} + \frac{\partial ^2 u}{\partial x \partial y} + \frac{\partial u}{\partial z} + u = 0 ∂t∂u+∂x∂y∂2u+∂z∂u+u=0
术语:
- order : n, 微分方程的最大阶数
- degree : r, 微分方程order的的次数(pow)
y
′
′
+
x
(
y
′
)
3
−
x
y
y'' + x(y')^3 - xy
y′′+x(y′)3−xy 是二阶(order=2)一次(degree=1)
y
′
′
=
y
′
+
y
+
x
y'' = \sqrt{y' + y + x}
y′′=y′+y+x是二阶二次方程
线性ODE:
n阶线性方程
∑
i
=
0
n
a
i
(
x
)
y
(
i
)
(
x
)
=
g
(
x
)
\sum_{i=0}^{n} a_i(x)y^{(i)}(x) = g(x)
i=0∑nai(x)y(i)(x)=g(x)
所有微分都是1次,系数是x的函数(a(x)和g(x)不要求是1次,可以是常数,但不是y的函数)
对函数
y
=
f
(
x
,
c
1
,
c
2
,
⋯
,
c
n
)
y = f(x, c_1, c_2, \cdots, c_n)
y=f(x,c1,c2,⋯,cn)逐次,得到n个方程,
联合n+1个方程消去
c
1
,
c
2
,
⋯
,
c
n
c_1, c_2, \cdots, c_n
c1,c2,⋯,cn, 得到
f
(
x
,
y
,
y
′
,
y
′
′
,
⋯
,
y
(
n
)
)
=
0
f(x, y, y', y'', \cdots, y^{(n)}) = 0
f(x,y,y′,y′′,⋯,y(n))=0
反过来理解: y = f ( x , c 1 , c 2 , ⋯ , c n ) y = f(x, c_1, c_2, \cdots, c_n) y=f(x,c1,c2,⋯,cn)是 f ( x , y , y ′ , y ′ ′ , ⋯ , y ( n ) ) = 0 f(x, y, y', y'', \cdots, y^{(n)}) = 0 f(x,y,y′,y′′,⋯,y(n))=0的通解
3.1.1 Initial & Boundary Value Problems 初始值问题和边界值问题
从通解得到特解
-
Initial Value Problem (IVP)
y ( i ) ( x 0 ) = α i , w h e r e i = 0 , 1 , ⋯ , n − 1 y^{(i)}(x_0)= \alpha_i, \ where \ i = 0,1, \cdots, n-1 y(i)(x0)=αi, where i=0,1,⋯,n−1,一共n个初始值条件
y ′ ′ + 2 y ′ = e x ; y ( π ) = 1 , y ′ ( π ) = 2 y'' + 2y' = e^x; \ y(\pi)=1, \ y'(\pi)=2 y′′+2y′=ex; y(π)=1, y′(π)=2 -
Boudary Value Problem (BVP)
y ( x i ) = β i , w h e r e i = 0 , 1 , ⋯ , n − 1 y(x_i)= \beta_i, \ where \ i = 0,1, \cdots, n-1 y(xi)=βi, where i=0,1,⋯,n−1,一共n个边界值条件
y ′ ′ + 2 y ′ = e x ; y ( 0 ) = 1 , y ( 1 ) = 1 y'' + 2y' = e^x; \ y(0)=1, \ y(1)=1 y′′+2y′=ex; y(0)=1, y(1)=1
3.2 First Order Ordinary Differential Equations 一阶常微分方程
一阶ODE的一般形式 F ( x , y , y ′ ) = 0 F(x, y, y')=0 F(x,y,y′)=0
3.2.1 One Variable Missing 缺少一个变量
x, y, y’中,没有x或者y的方程,可以用积分求解
-
没有y
y ′ = f ( x ) → y = ∫ f ( x ) d x y'=f(x) \rightarrow y = \int f(x)dx y′=f(x)→y=∫f(x)dx -
没有y’
y ′ = f ( y ) → x = ∫ 1 f ( y ) d x y'=f(y) \rightarrow x = \int \frac{1}{f(y)}dx y′=f(y)→x=∫f(y)1dx
3.2.2 Variable Separable 变量分离
通过积分得到通解
y ′ = g ( x ) h ( y ) → ∫ 1 h ( y ) d y = ∫ g ( x ) d x + c y' = g(x)h(y) \rightarrow \int \frac{1}{h(y)}dy = \int g(x)dx + c y′=g(x)h(y)→∫h(y)1dy=∫g(x)dx+c
3.2.3 Linear Equations 线性方程
用乘法法则进行组合,通过积分求解
求解
y
′
+
P
(
x
)
y
=
Q
(
x
)
→
R
(
x
)
y
′
+
R
(
x
)
P
(
x
)
y
=
d
d
x
(
R
(
x
)
y
)
=
R
(
x
)
Q
(
x
)
→
R
(
x
)
y
=
∫
R
(
x
)
Q
(
x
)
d
x
+
c
→
y
=
1
R
(
x
)
(
∫
R
(
x
)
Q
(
x
)
d
x
+
c
)
\begin{aligned} & y' + P(x)y = Q(x) \\ & \rightarrow R(x)y' + R(x)P(x)y = \frac{d}{dx}(R(x)y) = R(x)Q(x) \\ & \rightarrow R(x)y = \int R(x)Q(x)dx + c \\ & \rightarrow y = \frac{1}{R(x)}(\int R(x)Q(x)dx + c) \\ \end{aligned}
y′+P(x)y=Q(x)→R(x)y′+R(x)P(x)y=dxd(R(x)y)=R(x)Q(x)→R(x)y=∫R(x)Q(x)dx+c→y=R(x)1(∫R(x)Q(x)dx+c)
R(x)称为积分因子 Integrate Factor
R
(
x
)
Q
(
x
)
=
R
′
(
x
)
R
(
x
)
=
e
∫
P
(
x
)
d
x
R(x)Q(x)= R'(x) \\ R(x) = e^{\int P(x)dx}
R(x)Q(x)=R′(x)R(x)=e∫P(x)dx
3.3 Second Order ODE.s 二阶ODE
y ′ ′ = f ( x , y , y ′ ) y'' = f(x, y, y') y′′=f(x,y,y′)
3.3.1 Simplest Cases 简单情况
-
只有y’’, 没有y和y’:两次积分可以得到y
y ′ ′ = f ( x ) → y = ∫ ∫ f ( x ) d x y'' = f(x) \rightarrow y= \int \int f(x)dx y′′=f(x)→y=∫∫f(x)dx -
没有y:用P表示y’,用P’表示y’’,求解一阶ODE得到y’’
y ′ ′ = f ( y ′ , x ) y'' = f(y', x) y′′=f(y′,x)
令 P = y ′ → y ′ ′ = P ′ = f ( P , x ) P=y' \rightarrow y''=P'=f(P,x) P=y′→y′′=P′=f(P,x)
得到一阶ODE得到P,积分得到y -
没有y’和x:用p表示y’, 链式求导得到y’’(y),求解一阶ODE得到p(y)
y ′ ′ = f ( y ) y''=f(y) y′′=f(y)
令 P = y ′ → y ′ ′ = P ′ = d p d y d y d x = p d p d y = f ( y ) P=y' \rightarrow y'' = P' = \frac{dp}{dy} \frac{dy}{dx} = p \frac{dp}{dy} = f(y) P=y′→y′′=P′=dydpdxdy=pdydp=f(y)
得到一阶ODE的分离形式,两边积分得到f(y’, y)=0形式的一阶ODE,再次求解 -
没有x:用p表示y’, 链式求导得到y’’(y),求解一阶ODE
y ′ ′ = f ( y ′ , y ) y'' = f(y', y) y′′=f(y′,y)
令 P = y ′ → y ′ ′ = P ′ = d p d y d y d x = p d p d y = f ( P , y ) P = y' \rightarrow y'' = P' = \frac{dp}{dy} \frac{dy}{dx} = p \frac{dp}{dy} = f(P,y) P=y′→y′′=P′=dydpdxdy=pdydp=f(P,y)
得到一阶ODE的第三种形式
3.3.2 Linear ODE.s of Order at least 2 二次及二次以上ODE
n阶线性方程的一般形式
∑
i
=
0
n
a
i
(
x
)
y
(
i
)
(
x
)
=
g
(
x
)
\sum_{i=0}^{n} a_i(x)y^{(i)}(x) = g(x)
i=0∑nai(x)y(i)(x)=g(x)
n阶线性微分算子
D
≡
d
d
x
L
=
∑
i
=
0
n
a
i
D
i
L
y
=
g
(
x
)
D \equiv \frac{d}{dx} \\ L = \sum_{i=0}^{n} a_i D^i \\ L \ y = g(x) \\
D≡dxdL=i=0∑naiDiL y=g(x)
L是线性操作符
Homogeneous
如果g(x) = 0, 则
L
y
=
0
L \ y = 0
L y=0称为Homogeneous
L
y
=
g
(
x
)
L \ y = g(x)
L y=g(x)的通解分为两部分
y
=
y
c
+
y
p
y = y_c + y_p
y=yc+yp
y
c
y_c
yc称为 complimentary Function,是
L
y
=
0
L \ y = 0
L y=0的解
y
p
y_p
yp称为Paritcular Integral,是
L
y
=
g
(
x
)
L \ y = g(x)
L y=g(x)的解
3.3.3 Linear ODE.s with Constant Coeffcients 常系数线性ODE
考虑Homogeneous情况:
L
y
=
0
L \ y = 0
L y=0
一阶
L
y
=
a
y
′
+
b
y
=
0
L \ y = ay' + by=0
L y=ay′+by=0
二阶
L
y
=
a
y
′
′
+
b
y
′
+
c
y
=
0
L \ y = ay'' + by' + cy = 0
L y=ay′′+by′+cy=0
一阶方程的解为
y
=
e
λ
x
y = e^{\lambda x}
y=eλx
可以把一阶看作二阶的一个特例:也就是一阶方程的解应该也适用于二阶
将一阶方程的解代入二阶方程:
L
(
e
λ
x
)
=
(
a
λ
2
+
b
λ
+
c
)
e
λ
x
=
0
L(e^{\lambda x}) = (a \lambda ^2 + b \lambda + c)e^{\lambda x} = 0
L(eλx)=(aλ2+bλ+c)eλx=0
⟹
a
λ
2
+
b
λ
+
c
=
0
\Longrightarrow a \lambda ^2 + b \lambda + c = 0
⟹aλ2+bλ+c=0 称为AUXILLIARY EQUATION (A.E)
-
b 2 − 4 a c > 0 b^2-4ac > 0 b2−4ac>0 :
y = c 1 e λ 1 x + c 2 e λ 2 x y=c_1 e^{\lambda_1 x} + c_2 e^{\lambda_2 x} y=c1eλ1x+c2eλ2x -
b 2 − 4 a c = 0 b^2-4ac = 0 b2−4ac=0 :
除了 e λ x e^{\lambda x} eλx之外, x e λ x x e^{\lambda x} xeλx也是一个解
y = c 1 e λ x + c 2 x e λ x y=c_1 e^{\lambda x} + c_2 x e^{\lambda x} y=c1eλx+c2xeλx -
b 2 − 4 a c < 0 b^2-4ac < 0 b2−4ac<0 :
y = e p x ( A c o s q x + B s i n q x ) y = e^{px}(Acos \ qx + Bsin \ qx) y=epx(Acos qx+Bsin qx)
p是实部,q是虚部- 看推导过程,要求A是实数,B是纯虚数,
- 但是用通解代入时发现AB可以是任意数,不要求B为纯虚数
- 由于推导过程使用的常数项 C 1 , C 2 C_1, \ C_2 C1, C2并不要求是实数,因此B可以不是纯虚数
- 同样A也可以是虚数
3.4 General nth Order Equation n阶方程
L y = ∑ a n y ( n ) ≡ 0 L = ∑ a n D n A . E . ⟹ ∑ a n λ n = 0 L\ y = \sum a_n y^{(n)} \equiv 0 \\ L = \sum a_n D^{n} \\ A.E. \Longrightarrow \sum a_n \lambda ^ n = 0 L y=∑any(n)≡0L=∑anDnA.E.⟹∑anλn=0
A.E.解的种类
-
n个不同实数解
y = ∑ i = 1 n β i e λ i x y = \sum _{i=1}^{n} \beta _i e ^{\lambda_i x} y=i=1∑nβieλix -
有n重实数根 λ \lambda λ
y = e λ x ∑ i n α i x i − 1 y = e^{\lambda x} \sum_{i}^{n} \alpha_i x^{i - 1} y=eλxi∑nαixi−1 -
有n重共轭复数根 p ± i q p \pm iq p±iq
e p x [ ( ∑ i = 1 n A i x i − 1 ) c o s q x + ( ∑ i = 1 n B i x i − 1 ) s i n q x ] e^{px}[(\sum_{i=1}^{n} A_i x^{i-1})cos \ qx + (\sum_{i=1}^{n} B_i x^{i-1})sin \ qx] epx[(i=1∑nAixi−1)cos qx+(i=1∑nBixi−1)sin qx] -
以上三种情况的混合
3.5 Non-Homogeneous Case - Method of Undetermined Coefficients 非齐次
L y = g ( x ) , g ( x ) ≠ 0 L \ y = g(x), \ g(x) \neq 0 L y=g(x), g(x)=0
L y = L ( y c + y p ) = L y c + L y p L \ y = L \ (y_c + y_p) \\ = L \ y_c + L \ y_p L y=L (yc+yp)=L yc+L yp
通解: y = C . F . + P . I y = C.F. + P.I y=C.F.+P.I
齐次项
L
y
c
L \ y_c
L yc的通解
C
.
F
C.F
C.F 可以通过 A.E方程求解
非齐次项
L
y
p
L \ y_p
L yp的特解
P
.
I
P.I
P.I 可以通过以下方法求解
- “Guesswork” - which we are interested in
- Annihilator
- D-operator Method
Guesswork Method,猜测
g(x)的形式:
-
多项式
g ( x ) = ∑ i = 0 m p i x i g(x) = \sum_{i=0}^{m} p_i x^{i} g(x)=∑i=0mpixi
假设P.I的形式为 ∑ i = 0 m p i x i \sum_{i=0}^{m} p_i x^{i} ∑i=0mpixi, 与g(x)的次数相同,代入方程求解系数,可得到 y p y_p yp -
指数函数
g ( x ) = C e k x g(x) = Ce^{kx} g(x)=Cekx, 其中k不是A.E的解
假设P.I的形式为 A e k x Ae^{kx} Aekx, 代入方程求解A,可得到 y p y_p yp -
三角函数
g ( x ) = A s i n a x + B c o s a x g(x) = Asin\ ax + Bcos\ ax g(x)=Asin ax+Bcos ax, 其中ia不是A.E的解
假设特解的形式为 R e ( K e i a ) Re(K e ^ {ia}) Re(Keia), 代入方程求解K,可得到 y p y_p yp(取实部)
或者直接假设特解的形式为 A c o s a x + B s i n a x Acos\ ax + Bsin\ ax Acos ax+Bsin ax,求解A和B -
以上三种的组合
3.5.1 Failure Case 不可直接求解的形式
g ( x ) = C e k x g(x) = Ce^{kx} g(x)=Cekx, 其中k是A.E的根,
- 假设解为 A e k x Ae^{kx} Aekx,代入方程后,左边为0,无法求解
- 假设解为 A x r e k x Ax^re^{kx} Axrekx, r表示k是A.E的r重根(低次项已经包含齐次项通解中,不需要再求系数,也无法求解)
3.6 Linear ODE.s with Variable Coefficients - Euler Equation 可变系数的线性ODE - (柯西-)欧拉方程
Cauchy-Euler equation 柯西-欧拉方程:系数中x的次数,与微分的阶数相同
L
y
=
∑
i
=
0
n
a
i
x
i
y
(
i
)
=
g
(
x
)
L \ y = \sum_{i = 0}^{n} a_i x^i y^{(i)} = g(x)
L y=i=0∑naixiy(i)=g(x)
考虑二阶的情况,先求解齐次项,假设
y
=
x
λ
y = x^{\lambda}
y=xλ,代入方程得到
a
λ
2
+
(
b
−
a
)
λ
+
c
=
0
a \lambda^2 + (b-a) \lambda + c = 0
aλ2+(b−a)λ+c=0,三种情况:
- 两个实数解: y = A x λ 1 + B x λ 2 y = Ax^{\lambda_1} + Bx^{\lambda_2} y=Axλ1+Bxλ2
- 二重根: y = x λ ( A + B l n x ) y = x^{\lambda}(A + Bln \ x) y=xλ(A+Bln x)
- 共轭虚数根
y
=
x
α
(
A
c
o
s
(
β
l
n
x
)
+
B
s
i
n
(
β
l
n
x
)
)
y = x^{\alpha}(Acos(\beta ln\ x) + Bsin(\beta ln\ x))
y=xα(Acos(βln x)+Bsin(βln x))
x ± i β = e l n ( x ± i β ) = e ± i β l n x = e ± i θ θ = β l n x x^{\pm i \beta } = e ^ { ln(x^{\pm i \beta}) } = e^{\pm i \beta ln \ x} = e^{\pm i \theta} \\ \theta = \beta ln \ x x±iβ=eln(x±iβ)=e±iβln x=e±iθθ=βln x
3.6.1 Reduction to constant coefficient 退化为常系数方程
令
x
=
e
t
,
t
=
l
n
x
y
′
=
d
y
d
t
×
d
t
d
x
=
1
x
×
d
t
d
x
y
′
′
=
1
x
2
(
d
2
y
d
t
2
−
d
y
d
t
)
令x = e^t, \ t = ln \ x \\ y' = \frac{dy}{dt} \times \frac{dt}{dx} = \frac{1}{x} \times \frac{dt}{dx} \\ y'' = \frac{1}{x^2}(\frac{d^2 y}{dt^2} - \frac{dy}{dt})
令x=et, t=ln xy′=dtdy×dxdt=x1×dxdty′′=x21(dt2d2y−dtdy)
代入柯西-欧拉方程,刚好可以消去可变项,得到常系数方程
按常系数方程求解通解和特解,再用 t = l n x t = ln \ x t=ln x替换
3.7 Partial Differential Equations 偏微分方程(略)
1 Probability 概率
1.1 Preliminaries 知识准备
- experiment 实验:可重复,可产生一组输出的过程
- event 事件:一个或多个输出的集合
- sample space 样本空间:所有输出的集合,记作 Ω \Omega Ω
例子:掷六面色子
样本空间
Ω
=
{
1
,
2
,
3
,
4
,
5
,
6
}
\Omega = \{1, 2, 3, 4, 5, 6\}
Ω={1,2,3,4,5,6}
事件
- 掷出偶数: E 1 = { 2 , 4 , 6 } E_1 = \{2, 4, 6\} E1={2,4,6}
- 掷出奇数: E 1 = { 1 , 3 , 5 } E_1 = \{1, 3, 5\} E1={1,3,5}
- 掷出素数: E 1 = { 2 , 3 , 5 } E_1 = \{2, 3, 5\} E1={2,3,5}
1.1.1 Probability Scale 概率度量
事件E发生的概率P(E)满足条件 0 ≤ P ( E ) ≤ 1 0 \le P(E) \le 1 0≤P(E)≤1
1.1.2 Probability of an Event 事件概率
事件E发生的概率定义为
P ( E ) = 事 件 E 发 生 的 次 数 所 有 输 出 发 生 的 总 次 数 P(E) = \frac{事件E发生的次数}{所有输出发生的总次数} P(E)=所有输出发生的总次数事件E发生的次数
1.1.3 The Complimentary Event E’ 事件E的补集
E’表示E不发生,概率为 P ( E ′ ) = 1 − P ( E ) P(E') = 1 - P(E) P(E′)=1−P(E)
1.2 Probability Diagrams 用图表表示概率
三种表示方法
-
样本空间或二维表格
用两个色子掷出8的概率 P ( 8 ) = 5 36 P(8)=\frac{5}{36} P(8)=365
-
树型图
4红,5黄,11蓝,每次取出后放回,取两次,取出1个红球,1个蓝球的概率
4 20 × 11 20 + 11 20 × 4 20 = 11 50 \frac{4}{20} \times \frac{11}{20} + \frac{11}{20} \times \frac{4}{20} = \frac{11}{50} 204×2011+2011×204=5011
-
Venn Diagram 文氏图:计算并集、交集、条件概率
1.3 Conditional Probability 条件概率
给定事件A发生的情况下,事件B发生的概率。记作 P ( B ∣ A ) P(B|A) P(B∣A)
P ( A ∣ B ) = P ( A ∩ B ) P ( B ) P ( A ∩ B ) = P ( A ∣ B ) × P ( B ) P(A|B)=\frac{P(A \cap B)}{P(B)} \\ P(A \cap B) = P(A|B) \times P(B) P(A∣B)=P(B)P(A∩B)P(A∩B)=P(A∣B)×P(B)
Bayes’ Theorem
P ( A ∩ B ) = P ( B ∩ A ) P ( A ∣ B ) × P ( B ) = P ( B ∣ A ) × P ( A ) P ( B ∣ A ) = P ( A ∣ B ) × P ( B ) P ( A ) P(A \cap B) = P(B \cap A) \\ P(A|B) \times P(B) = P(B|A) \times P(A) \\ P(B|A) = \frac{P(A|B) \times P(B)}{P(A)} P(A∩B)=P(B∩A)P(A∣B)×P(B)=P(B∣A)×P(A)P(B∣A)=P(A)P(A∣B)×P(B)
例子:10个硬币,9个普通硬币(一正一反),1个特殊硬币(两个正面),从中取出一个硬币,放5次,问题
-
取到5个正面的概率
取到普通硬币的概率 P ( N ) = 9 10 P(N) = \frac{9}{10} P(N)=109
取到普通硬币后,得到正面的概率 P ( H ∣ N ) = 1 2 P(H|N) = \frac{1}{2} P(H∣N)=21
取到特殊硬币的概率 P ( S ) = 1 − P ( N ) = 1 10 P(S) = 1 - P(N) = \frac{1}{10} P(S)=1−P(N)=101
取到特殊硬币后,得到正面的概率 P ( H ∣ S ) = 1 P(H|S) = 1 P(H∣S)=1
P ( 5 H e a d s ) = 9 10 × ( 1 2 ) 5 + 1 10 × 1 5 = 41 320 P(5 \ Heads) = \frac{9}{10} \times (\frac{1}{2})^5 + \frac{1}{10} \times 1^5 = \frac{41}{320} P(5 Heads)=109×(21)5+101×15=32041 -
取到5个正面时,这个硬币是特殊硬币的概率
P ( S ∣ 5 H e a d s ) = P ( 5 H e a d s ∣ S ) × P ( S ) P ( 5 H e a d s ) = 1 × 1 10 41 320 = 32 41 P(S | 5 \ Heads) = \frac{P(5 \ Heads | S) \times P(S)}{P(5 \ Heads)} = \frac{1 \times \frac{1}{10}}{\frac{41}{320}} = \frac{32}{41} P(S∣5 Heads)=P(5 Heads)P(5 Heads∣S)×P(S)=320411×101=4132
1.4 Mutually exclusive and Independent events 互斥和独立
互斥:
P
(
A
∩
B
)
=
0
P(A \cap B) = 0
P(A∩B)=0
P
(
A
∪
B
)
=
P
(
A
)
+
P
(
B
)
P(A \cup B) = P(A) + P(B)
P(A∪B)=P(A)+P(B)
独立:
P
(
A
∩
B
)
=
P
(
A
)
×
P
(
B
)
P(A \cap B) = P(A) \times P(B)
P(A∩B)=P(A)×P(B)
P
(
A
∣
B
)
=
P
(
A
)
P(A|B) = P(A)
P(A∣B)=P(A)
1.5 Two famous problems
Birthday Problem:60个人中至少两个人是同一个生日的概率
所有人生日都不相同的概率
P
=
365
×
364
×
⋯
×
306
36
5
60
P = \frac{365 \times 364 \times \cdots \times 306}{365 ^ {60}}
P=36560365×364×⋯×306
有至少两个人生日相同的概率为
1
−
P
=
0.5
1-P = 0.5%
1−P=0.5
Monty Hall Game Show:ABC三个门,其中一个里面有车,另外两个有山羊,甲随机选了一个门,主持人去掉一个有山羊的门,这时候甲应该改变选择吗?
问题的关键在于主持人知道哪个门后面有车
1.6 Random Variables 随机变量
1.6.1 Notation 术语
随机变量:
X
,
Y
,
Z
X, \ Y, \ Z
X, Y, Z
观察变量:
x
,
y
,
z
x, \ y, \ z
x, y, z
1.6.2 Definition 定义
值空间不一定都是数值,进行从值空间到数轴的映射
1.6.3 Types of Random variable 随机变量的种类
- 离散:值可数
- 连续:值不可数
1.7 Probability Distributions 概率分布
1.7.1 Discrete distributions 离散分布
同时掷两个色子,和的概率分布
1.7.2 Continuous Distributions 连续分布
单个值的概率分布为0 P ( X = x ) = 0 P(X=x)=0 P(X=x)=0
用PDF描述概率:Probablity Density Function
P
(
a
<
X
<
b
)
=
∫
a
b
f
(
x
)
d
x
P(a \lt X \lt b) = \int _a^b f(x) dx
P(a<X<b)=∫abf(x)dx
PDF的特性:
- f ( x ) ≥ 0 f(x) \ge 0 f(x)≥0
- ∫ − ∞ ∞ f ( x ) d x = 1 \int _{-\infty}^{\infty} f(x) dx = 1 ∫−∞∞f(x)dx=1
- P ( a < X < b ) = ∫ a b f ( x ) d x P(a \lt X \lt b) = \int _a^b f(x) dx P(a<X<b)=∫abf(x)dx
1.8 Cumulative Distribution Function CDF 累积分布函数
概率累积: F ( x ) = P ( X ≤ x ) F(x) = P(X \le x) F(x)=P(X≤x)
1.8.1 Discrete Random variables
对小于x的值的概率进行累加
1.8.2 Continuous Random variables 连续随机变量
F ( x ) = P ( X ≤ x ) = ∫ − ∞ x f ( t ) d t F(x) = P(X \le x) = \int_{-\infty}^{x} f(t)dt F(x)=P(X≤x)=∫−∞xf(t)dt
1.9 Expectation and Variance 期望和方差
期望:
E
(
X
)
=
μ
E(X) = \mu
E(X)=μ,描述分布均值
方差:
V
(
X
)
=
σ
2
V(X) = \sigma^2
V(X)=σ2,描述分布离散程度
1.9.1 Discrete Random variables 离散随机变量
E
(
X
)
=
∑
a
l
l
x
x
P
(
X
=
x
)
E(X) = \sum_{all \ x} xP(X=x)
E(X)=∑all xxP(X=x)
V
(
X
)
=
E
(
X
−
E
(
X
)
)
=
E
(
X
2
)
−
(
E
(
X
)
)
2
V(X) = E(X - E(X)) = E(X^2) - (E(X))^2
V(X)=E(X−E(X))=E(X2)−(E(X))2
1.9.2 Continuous Random Variables 连续随机变量
E
(
X
)
=
∫
a
l
l
x
x
f
(
x
)
d
x
E(X) = \int_{all \ x} xf(x)dx
E(X)=∫all xxf(x)dx
V
(
X
)
=
E
[
(
X
−
E
(
X
)
)
2
]
=
E
(
X
2
)
−
(
E
(
X
)
)
2
=
∫
a
l
l
x
x
2
f
(
x
)
d
x
−
(
∫
a
l
l
x
x
f
(
x
)
d
x
)
2
V(X) = E[(X - E(X))^2] = E(X^2) - (E(X))^2 = \int_{all \ x} x^2 f(x)dx - (\int_{all \ x} xf(x)dx)^2
V(X)=E[(X−E(X))2]=E(X2)−(E(X))2=∫all xx2f(x)dx−(∫all xxf(x)dx)2
1.10 Expectation Algebra 期望值的代数运算
期望是线性操作
E
(
X
+
a
)
=
E
(
X
)
+
a
E
(
a
X
)
=
a
E
(
X
)
E
(
X
+
Y
)
=
E
(
X
)
+
E
(
Y
)
E(X + a) = E(X) +a \\ E(aX) = aE(X) \\ E(X + Y) = E(X) + E(Y) \\
E(X+a)=E(X)+aE(aX)=aE(X)E(X+Y)=E(X)+E(Y)
V
(
X
+
a
)
=
V
(
X
)
V
(
a
X
)
=
a
2
V
(
X
)
V
(
b
)
=
0
V(X + a) = V(X) \\ V(aX) = a^2V(X) \\ V(b) = 0 \\
V(X+a)=V(X)V(aX)=a2V(X)V(b)=0
如果X和Y独立
E
(
X
Y
)
=
E
(
X
)
E
(
Y
)
V
(
X
+
Y
)
=
V
(
X
)
+
V
(
Y
)
E(XY) = E(X)E(Y) \\ V(X + Y) = V(X) + V(Y) \\
E(XY)=E(X)E(Y)V(X+Y)=V(X)+V(Y)
如果不独立
V
(
X
+
Y
)
=
V
(
X
)
+
V
(
Y
)
+
2
E
[
(
X
−
E
(
X
)
)
(
Y
−
E
(
Y
)
]
C
o
v
(
X
,
Y
)
=
E
[
(
X
−
E
(
X
)
)
(
Y
−
E
(
Y
)
]
V(X + Y) = V(X) + V(Y) + 2E[(X-E(X))(Y-E(Y)] \\ Cov(X, Y) = E[(X-E(X))(Y-E(Y)]
V(X+Y)=V(X)+V(Y)+2E[(X−E(X))(Y−E(Y)]Cov(X,Y)=E[(X−E(X))(Y−E(Y)]
1.11 Moments 矩
一阶矩:
E
(
X
)
=
μ
E(X) = \mu
E(X)=μ
n阶矩:
E
(
X
n
)
=
∫
a
l
l
x
x
n
f
(
x
)
d
x
E(X^n) = \int_{all \ x} x^n f(x) dx
E(Xn)=∫all xxnf(x)dx
中心距:
-
二阶中心矩:方差 V ( X ) = E [ ( X − E ( X ) ) 2 ] = σ 2 V(X) = E[(X - E(X)) ^ 2] = \sigma^2 V(X)=E[(X−E(X))2]=σ2
-
三阶中心距: E [ ( X − E ( X ) ) 3 ] E[(X-E(X))^3] E[(X−E(X))3]
-
偏度(三阶归一化中心距):度量非对称性
S k e w n e s s = E [ ( X − E ( X ) ) 3 ] σ 3 Skewness = \frac {E[(X-E(X)) \ ^3]}{\sigma ^ 3} Skewness=σ3E[(X−E(X)) 3] -
尾度(四阶归一化中心距):度量肥尾特性
K u r t o s i s = E [ ( X − E ( X ) ) 4 ] σ 4 Kurtosis= \frac {E[(X-E(X)) \ ^4]}{\sigma ^ 4} Kurtosis=σ4E[(X−E(X)) 4]- kurtosis越大,表示尾部数据的概率分布越高(fat tail 肥尾)
- 正态分布的kurtosis为3
- Leptokurtic: kurtosis大于3,PDF图像呈现尖峰肥尾
- Platykurtic: kurtosis小于3,PDF图像比较平坦
1.12 Covariance 协方差
如果X和Y独立 E ( X Y ) = E ( X ) E ( Y ) E(XY) = E(X)E(Y) E(XY)=E(X)E(Y)$
协方差: C o v ( X , Y ) = E [ ( X − E ( X ) ) ( Y − E ( Y ) ] = E ( X Y ) − E ( X ) E ( Y ) Cov(X, Y) = E[(X-E(X))(Y-E(Y)] = E(XY) - E(X)E(Y) Cov(X,Y)=E[(X−E(X))(Y−E(Y)]=E(XY)−E(X)E(Y)
相关性:
ρ
X
Y
=
C
o
v
(
X
,
Y
)
σ
X
σ
Y
\rho _{XY} = \frac{Cov(X, Y)}{\sigma_X \sigma_Y}
ρXY=σXσYCov(X,Y)
ρ X Y \rho _{XY} ρXY 等于-1,负相关,反向变化,等于1,正相关,同向变化,等于0,不相关
1.13 Important Distributions 重要分布
这些概率分布适用于对不同的对象进行建模
1.13.1 Binomial Distribution 二项式分布
多次伯努利实验的累积
- 实验次数为n
- 实验之间相互独立
- 成功概率(=1的概率)为p
X ∼ B ( n , p ) P ( X = x ) = C x n p x ( 1 − p ) n − x C x n = n ! x ! ( n − x ) ! E ( X ) = n p V ( X ) = n p ( 1 − p ) \begin{aligned} X & \sim B(n, p) \\ P(X=x) & = C_x^n p^x(1-p)^{n-x} \\ C_x^n& = \frac{n!}{x!(n-x)!} \\ E(X) & = np \\ V(X) & = np(1-p) \end{aligned} XP(X=x)CxnE(X)V(X)∼B(n,p)=Cxnpx(1−p)n−x=x!(n−x)!n!=np=np(1−p)
1.13.2 Poisson Distribution 泊松分布
泊松分布:对违约个数进行建模
- 两个事件同时发生时,没有影响
- 事件之间相互独立
- 事件在所有时间点发生的概率相同
X ∼ P o ( λ ) P ( X = r ) = e − λ λ r r ! , w h e r e r = 0 , 1 , 2 , ⋯ E ( X ) = λ V ( X ) = λ \begin{aligned} X & \sim Po(\lambda) \\ P(X=r) & = \frac{e^{-\lambda}\lambda^r} {r!}, \ where \ r = 0, 1, 2, \cdots \\ E(X) & = \lambda \\ V(X) & = \lambda \end{aligned} XP(X=r)E(X)V(X)∼Po(λ)=r!e−λλr, where r=0,1,2,⋯=λ=λ
1.13.3 Normal Distribution 正态分布
正态分布时最常用连续分布
X ∼ N ( μ , σ 2 ) E ( X ) = μ V ( X ) = σ 2 P D F = f ( x ) = 1 σ 2 π e − ( x − μ ) 2 2 σ 2 C D F = P ( X ≤ x ) = ∫ − ∞ x 1 σ 2 π e − ( s − μ ) 2 2 σ 2 d s ∫ − ∞ ∞ 1 σ 2 π e ( x − μ ) 2 2 σ 2 d x = 1 \begin{aligned} X & \sim N(\mu, \sigma^2) \\ E(X) & = \mu \\ V(X) & = \sigma^2 \\ PDF=f(x) & = \frac{1}{\sigma\sqrt{2\pi}} e^{- \frac{(x-\mu)^2}{2\sigma^2} } \\ CDF=P(X \leq x) & = \int_{-\infty}^{x}\frac{1}{\sigma\sqrt{2\pi}}e^{-\frac{(s-\mu)^2}{2\sigma^2}}ds \\ \int_{-\infty}^{\infty}\frac{1}{\sigma\sqrt{2\pi}}e^{\frac{(x-\mu)^2}{2\sigma^2}}dx & = 1 \\ \end{aligned} XE(X)V(X)PDF=f(x)CDF=P(X≤x)∫−∞∞σ2π1e2σ2(x−μ)2dx∼N(μ,σ2)=μ=σ2=σ2π1e−2σ2(x−μ)2=∫−∞xσ2π1e−2σ2(s−μ)2ds=1
1.13.4 Standard Normal distribution 标准正态分布(z分布)
Z ∼ N ( 0 , 1 ) E ( X ) = 0 V ( X ) = 1 P D F = f ( z ) = 1 2 π e − z 2 2 C D F = Φ ( z ) = ∫ − ∞ z 1 2 π e − s 2 2 d s Z = X − μ σ \begin{aligned} Z & \sim N(0, 1) \\ E(X) & = 0 \\ V(X) & = 1 \\ PDF=f(z) & = \frac{1}{\sqrt{2\pi}}e^{- \frac{z^2}{2}} \\ CDF=\Phi(z) & = \int_{-\infty}^{z}\frac{1}{\sqrt{2\pi}}e^{- \frac{s^2}{2}}ds \\ Z & = \frac{X - \mu}{\sigma} \end{aligned} ZE(X)V(X)PDF=f(z)CDF=Φ(z)Z∼N(0,1)=0=1=2π1e−2z2=∫−∞z2π1e−2s2ds=σX−μ
1.13.5 Common regions 常用区间
正态分布的区间概率
区间 | 概率 |
---|---|
± σ \pm\sigma ±σ | 68% |
± 2 σ \pm2\sigma ±2σ | 95% |
± 3 σ \pm3\sigma ±3σ | 99.8% |
1.14 Central Limit Theorem 中心极限定理
X 1 , X 2 , ⋯ , X n X_1, X_2, \cdots, X_n X1,X2,⋯,Xn 是独立同分布(independent same distribution, iid)的随机变量
样本均值 X ˉ = ∑ i = 1 n X i n \bar X = \frac{\sum_{i=1}^{n}X_i}{n} Xˉ=n∑i=1nXi符合正态分布
- X的分布可以是任意分布
- 只需要很小的n就可以产生正态分布
- Total and means are quantities of interest
E ( X ˉ ) = μ V ( X ˉ ) = σ 2 / n X ˉ ∼ N ( μ , σ 2 ) X ˉ − μ σ n ∼ N ( 0 , 1 ) \begin{aligned} E(\bar X) & = \mu \\ V(\bar X) & = \sigma^2/n \\ \bar X & \sim N(\mu, \sigma^2) \\ \frac{\bar X - \mu}{\sigma \sqrt{n}} & \sim N(0, 1) \end{aligned} E(Xˉ)V(Xˉ)XˉσnXˉ−μ=μ=σ2/n∼N(μ,σ2)∼N(0,1)
2 Statistics 统计
2.1 Sampling 采样
采样的权重(概率)都是 1 N \frac{1}{N} N1
总体
μ
=
1
N
∑
i
=
1
N
x
i
σ
2
=
1
N
∑
i
=
1
N
(
x
i
−
μ
)
2
\begin{aligned} \mu & = \frac{1}{N}\sum_{i=1}^{N}x_i \\ \sigma^2 & = \frac{1}{N}\sum_{i=1}^{N}(x_i - \mu)^2 \end{aligned}
μσ2=N1i=1∑Nxi=N1i=1∑N(xi−μ)2
样本
X
ˉ
=
1
n
∑
i
=
1
n
x
i
E
(
X
ˉ
)
=
μ
S
2
=
1
n
−
1
∑
i
=
1
n
(
x
i
−
x
ˉ
)
2
E
(
s
2
)
=
σ
2
\begin{aligned} \bar X & = \frac{1}{n}\sum_{i=1}^{n}x_i \\ E(\bar X) & = \mu \\ S^2 & = \frac{1}{n-1}\sum_{i=1}^{n}(x_i - \bar x)^2 \\ E(s^2) & = \sigma^2 \end{aligned}
XˉE(Xˉ)S2E(s2)=n1i=1∑nxi=μ=n−11i=1∑n(xi−xˉ)2=σ2
2.1.1 Proof 证明
独立同分布假设可以得到 V ( ∑ i = 1 n x i ) = n σ 2 V(\sum_{i=1}^nx_i) = n\sigma^2 V(∑i=1nxi)=nσ2
根据中心极限定理,得到:
E
(
X
ˉ
)
=
μ
V
(
X
ˉ
)
=
σ
2
n
V
(
X
ˉ
)
=
E
(
X
ˉ
2
)
−
[
E
(
X
ˉ
)
]
2
⟹
E
(
X
ˉ
2
)
=
σ
2
n
+
μ
2
令
n
=
1
⟹
E
(
X
i
2
)
=
σ
2
+
μ
2
E
(
S
2
)
=
1
n
−
1
E
[
∑
(
X
i
−
X
ˉ
)
2
]
=
1
n
−
1
E
[
∑
X
i
2
−
n
X
ˉ
2
]
=
1
n
−
1
∑
E
(
X
i
2
)
−
n
[
E
(
X
ˉ
)
]
2
=
1
n
−
1
(
n
(
σ
2
+
μ
2
)
−
n
(
σ
2
n
+
μ
2
)
)
=
σ
2
\begin{aligned} E(\bar X) & = \mu \\ V(\bar X) & = \frac{\sigma ^ 2}{n} \\ V(\bar X) & = E(\bar X^2) - [E(\bar X)]^2 \Longrightarrow E(\bar X^2) & = \frac{\sigma^2}{n} + \mu^2 \\ 令n=1 \Longrightarrow E(X_i^2) & = {\sigma^2} + \mu^2 \\ E(S^2) & = \frac{1}{n-1}E[\sum(X_i - \bar X)^2] \\ & = \frac{1}{n-1}E[\sum X_i^2 - n \bar X^2] \\ & = \frac{1}{n-1}\sum E(X_i^2) - n[E(\bar X)]^2 \\ & = \frac{1}{n-1}(n({\sigma^2} + \mu^2) - n ( \frac{\sigma^2}{n} + \mu^2)) = \sigma ^2 \\ \end{aligned}
E(Xˉ)V(Xˉ)V(Xˉ)令n=1⟹E(Xi2)E(S2)=μ=nσ2=E(Xˉ2)−[E(Xˉ)]2⟹E(Xˉ2)=σ2+μ2=n−11E[∑(Xi−Xˉ)2]=n−11E[∑Xi2−nXˉ2]=n−11∑E(Xi2)−n[E(Xˉ)]2=n−11(n(σ2+μ2)−n(nσ2+μ2))=σ2=nσ2+μ2
2.2 Maximum Likelihood Estimation 极大似然估计 MLE
- 统计方法
- 用数据拟合模型
用观测值估计模型的参数:按照给定的数据,什么样的模型参数可能会给出这样的数据
2.2.1 Motivating example
- 先假设数据服从某个分布(带参数的分布),例如二项式分布
- 观测一组数据(独立同分布)
- 将数据代入似然函数(分布函数),求解使得似然函数取最大值的参数
求梯度得到极值对应的参数值
2.2.2 In General
l
(
θ
;
x
1
,
x
2
,
x
3
,
⋯
,
x
n
)
=
f
(
x
1
,
x
2
,
x
3
,
⋯
,
x
n
;
θ
)
L
(
θ
;
x
1
,
x
2
,
x
3
,
⋯
,
x
n
)
=
l
o
g
l
(
θ
;
x
1
,
x
2
,
x
3
,
⋯
,
x
n
)
\begin{aligned} l(\theta; x_1, x_2, x_3, \cdots, x_n) & = f(x_1, x_2, x_3, \cdots, x_n; \theta) \\ L(\theta; x_1, x_2, x_3, \cdots, x_n) & = log \ l(\theta; x_1, x_2, x_3, \cdots, x_n) \end{aligned}
l(θ;x1,x2,x3,⋯,xn)L(θ;x1,x2,x3,⋯,xn)=f(x1,x2,x3,⋯,xn;θ)=log l(θ;x1,x2,x3,⋯,xn)
目标:找到使得L最大的参数
θ
\theta
θ
2.2.3 Normal Distribution
估算 μ 和 σ \mu和\sigma μ和σ
假设
- X ∼ N ( μ , σ 2 ) X \sim N(\mu, \sigma^2) X∼N(μ,σ2), i.e. f ( x ) = 1 σ 2 π e − ( x − μ ) 2 2 σ 2 f(x)=\frac{1}{\sigma\sqrt{2\pi}}e^{-\frac{(x-\mu)^2}{2\sigma^2}} f(x)=σ2π1e−2σ2(x−μ)2
- x 1 , x 2 , x 3 , ⋯ , x n x_1, x_2, x_3, \cdots, x_n x1,x2,x3,⋯,xn是iid的随机采样
L ( μ , σ ; x 1 , x 2 , x 3 , ⋯ , x n ) = l o g l ( μ , σ ; x 1 , x 2 , x 3 , ⋯ , x n ) = l o g f ( x 1 , x 2 , x 3 , ⋯ , x n ; μ , σ ) = l o g ∏ i = 1 n f ( x i ; μ , σ ) = ∑ i = 1 n l o g f ( x i ; μ , σ ) = ∑ i = 1 n l o g ( 1 σ 2 π e − ( x i − μ ) 2 2 σ 2 ) = ∑ i = 1 n ( l o g ( 1 σ 2 π ) − ( x i − μ ) 2 2 σ 2 ) = − n 2 l o g ( 2 π ) − n l o g ( σ ) − 1 2 σ 2 ∑ i = 1 n ( x i − μ ) 2 \begin{aligned} & L(\mu, \sigma; x_1, x_2, x_3, \cdots, x_n) \\ = & log \ l(\mu, \sigma; x_1, x_2, x_3, \cdots, x_n) \\ = & log \ f(x_1, x_2, x_3, \cdots, x_n; \mu, \sigma) \\ = & log \prod_{i=1}^n f(x_i; \mu, \sigma) \\ = & \sum_{i=1}^n logf(x_i; \mu, \sigma) \\ = & \sum_{i=1}^n log(\frac{1}{\sigma\sqrt{2\pi}}e^{-\frac{(x_i-\mu)^2}{2\sigma^2}}) \\ = & \sum_{i=1}^n (log(\frac{1}{\sigma\sqrt{2\pi}})-\frac{(x_i-\mu)^2}{2\sigma^2}) \\ = & -\frac{n}{2}log(2\pi)-nlog(\sigma)-\frac{1}{2\sigma^2}\sum_{i=1}^n(x_i - \mu)^2 \end{aligned} =======L(μ,σ;x1,x2,x3,⋯,xn)log l(μ,σ;x1,x2,x3,⋯,xn)log f(x1,x2,x3,⋯,xn;μ,σ)logi=1∏nf(xi;μ,σ)i=1∑nlogf(xi;μ,σ)i=1∑nlog(σ2π1e−2σ2(xi−μ)2)i=1∑n(log(σ2π1)−2σ2(xi−μ)2)−2nlog(2π)−nlog(σ)−2σ21i=1∑n(xi−μ)2
对
μ
\mu
μ和
σ
\sigma
σ求偏微分,得到的解为
μ
=
1
n
∑
i
=
1
n
x
i
σ
=
1
n
∑
i
=
1
n
(
x
i
−
μ
)
2
\begin{aligned} \mu &= \frac{1}{n}\sum_{i=1}^n x_i \\ \sigma &= \frac{1}{n}\sum_{i=1}^n (x_i - \mu)^2 \end{aligned}
μσ=n1i=1∑nxi=n1i=1∑n(xi−μ)2
2.3 Regression and Correlation 回归和相关性
2.3.1 Linear regression 线性回归
两变量 y = a x + b y=ax+b y=ax+b, 计算系数a和b使得回归误差最小
2.3.2 Correlation 相关系数
2.3.3 Pearson Product-Moment Correlation Coefficient
P M C C r = s x y S x x S y y PMCC \ r = \frac{s_{xy}}{\sqrt{S_{xx}S_{yy}}} PMCC r=SxxSyysxy
2.3.4 Spearman’s Rank Correlation Coefficient 基于排序的相关性
2.4 Time Series 时间序列
两种方法对时间序列建模
- Simple moving average:MA 简单移动平均
- Auto Regressive :AR 自回归