机器学习之路 - 微积分

机器学习之路 - 微积分

持续更新中…

路漫漫其修远兮,机器学习之路从数学开始,包括概率论与数理统计,微积分基础,线性代数(矩阵运算),本文总结了微积分相关的概念和性质

1. 邻域

U ( a , δ ) ≜ { x   ∣    ∣ x − a ∣ < δ } = ( a − δ , a + δ ) U(a,\delta)\triangleq \{x\ |\ \ |x-a|<\delta\}=(a-\delta,a+\delta) U(a,δ){x   xa<δ}=(aδ,a+δ)
a

U ( a , δ ) U(a,\delta) U(a,δ)为a的 δ \delta δ邻域
U ˚ ( a , δ ) ≜ { x ∣ 0 < ∣ x − a ∣ < δ } \mathring{U}(a,\delta)\triangleq \{x|0<|x-a|<\delta\} U˚(a,δ){x0<xa<δ}

b

U ˚ ( a , δ ) \mathring{U}(a,\delta) U˚(a,δ)为a的去心 δ \delta δ邻域,左邻域,右邻域

2. 常用函数

  • 绝对值函数
    y = ∣ x ∣ = { − x   x < 0 x      x ≤ 0 y=|x|=\begin{cases} -x\:x\lt0 \\ x\:\:\:\:x\le0 \end{cases} y=x={xx<0xx0
    001

  • 符号函数
    y = { − 1   x < 0 0   x = 0 1   x > 0 y=\begin{cases} -1 \:x\lt0 \\ 0\:x=0 \\ 1\:x\gt0 \end{cases} y=1x<00x=01x>0
    002

  • 分段函数
    y = f ( x ) = { 2 x   0 ≤ x ≤ 1 1 + x   x > 1 y=f(x)=\begin{cases} 2\sqrt{x}\:0\le x\le 1 \\ 1+x\:x\gt1 \end{cases} y=f(x)={2x 0x11+xx>1
    003

  • 取整函数

    [ x ] ≜ [x]\triangleq [x]最接近 x x x左边的整数,即向下取整
    y = f ( x ) = [ x ] y=f(x)=[x] y=f(x)=[x]
    004

  • 狄利克雷(Dirichlet)函数
    D ( x ) = { 1   x ∈ Q 0   x ∈ Q c D(x)=\begin{cases} 1\: x\in Q \\ 0\:x\in Q^c \end{cases} D(x)={1xQ0xQc

3. 函数性质

  1. 有界性

    y = f ( x ) ( x ∈ D ) y=f(x)\quad(x\in D) y=f(x)(xD)

    ∃ M > 0 , ∀ x ∈ D \exist M>0, \forall x\in D M>0,xD,有 ∣ f ( x ) ∣ ≤ M |f(x)|\le M f(x)M

    f ( x ) f(x) f(x) D D D上有界

    f ( x ) ≥ M 1 f(x)\ge M_1 f(x)M1,则 f ( x ) f(x) f(x)有下界或

    c

    f ( x ) ≤ M 2 f(x)\le M_2 f(x)M2,则 f ( x ) f(x) f(x)有上界

  2. 单调性

    y = f ( x ) ( x ∈ D ) y=f(x)\quad(x\in D) y=f(x)(xD)

    ∀ x 1 , x 2 ∈ D 且 x 1 < x 2 \forall x_1,x_2\in D且x_1<x_2 x1,x2Dx1<x2,有 f ( x 1 ) < f ( x 2 ) f(x_1)<f(x_2) f(x1)<f(x2)

    y = f ( x ) y=f(x) y=f(x) x ∈ D x\in D xD上严格递减

    另有单调递增,单调不减,单调不增

  3. 反函数

    y = f ( x ) ( x ∈ D ) y=f(x)(x\in D) y=f(x)(xD)

    y = f ( x ) ⇒ x = ϕ ( y ) y=f(x)\Rightarrow x=\phi(y) y=f(x)x=ϕ(y)

    { x + x 2 + 1 = e y ① − x + x 2 + 1 = e − y ② \begin{cases}x+\sqrt{x^2+1}=e^y\qquad①\\ -x+\sqrt{x^2+1}=e^{-y}\qquad②\end{cases} {x+x2+1 =eyx+x2+1 =ey

    ① − ② ⇒ 2 x = e y − e − y ⇒ x = e y − e − y 2 ①-②\Rightarrow2x=e^y-e^{-y}\Rightarrow x=\frac{e^y-e^{-y}}{2} 2x=eyeyx=2eyey

    反函数为: x = e y − e − y 2 x=\frac{e^y-e^{-y}}{2} x=2eyey

  4. 奇偶性

    y = f ( x ) , x ∈ D , D 关 于 原 点 对 称 y=f(x),x\in D,D关于原点对称 y=f(x),xD,D

    { f ( − x ) = − f ( x ) ,   y = f ( x ) 为 奇 函 数 f ( − x ) = f ( x ) ,   y = f ( x ) 为 偶 函 数 \begin{cases} f(-x)=-f(x),\:y=f(x)为奇函数 \\ f(-x)=f(x),\:y=f(x)为偶函数 \end{cases} {f(x)=f(x),y=f(x)f(x)=f(x),y=f(x)

  5. 周期性: f ( x + l ) = f ( x ) f(x+l)=f(x) f(x+l)=f(x)恒成立

    l l l为周期,通常指最小正周期。 s i n   x sin\:x sinx c o s   x cos\:x cosx都是以 2 π 2\pi 2π为周期的周期函数, t a n   x tan\:x tanx是以 π \pi π为周期的周期函数

    在每个长度为 l l l的区间上,函数图形有相同形状。

4.极限

性 质 { 一 般 性 质 { 唯 一 性 保 号 性 局 部 有 界 性 运 算 性 质 存 在 性 质 性质\begin{cases} 一般性质\begin{cases} 唯一性 \\ 保号性 \\ 局部有界性 \end{cases} \\ 运算性质 \\ 存在性质 \end{cases}

4.1 数列极限

4.1.1 定义

a n = n n + 1 : 1 2 , 2 3 , 3 4 , … a_n=\frac{n}{n+1}:\frac{1}{2},\frac{2}{3},\frac{3}{4},\dots an=n+1n:21,32,43,

∣ a n − 1 ∣ = 1 n + 1 |a_n-1|=\frac{1}{n+1} an1=n+11

ϵ 1 = 1 10 \epsilon_1=\frac{1}{10} ϵ1=101, ∣ a n − 1 ∣ = 1 n + 1 < 10 ⇒ n > 9 |a_n-1|=\frac{1}{n+1}\lt 10\Rightarrow n\gt 9 an1=n+11<10n>9

ϵ 1 = 1 100 \epsilon_1=\frac{1}{100} ϵ1=1001, ∣ a n − 1 ∣ = 1 n + 1 < 100 ⇒ n > 99 |a_n-1|=\frac{1}{n+1}\lt 100\Rightarrow n\gt 99 an1=n+11<100n>99

… \dots

∀ ϵ > 0 , ∣ a n − 1 ∣ = 1 n + 1 < ϵ    ⟺    n > [ 1 ϵ ] − 1 , 即 lim ⁡ n → ∞ a n = 1 \forall \epsilon\gt 0,|a_n-1|=\frac{1}{n+1}\lt\epsilon\iff n\gt [\frac{1}{\epsilon}]-1,即\lim_{n\to\infin}a_n=1 ϵ>0,an1=n+11<ϵn>[ϵ1]1,limnan=1

a n = 1 n 2 + 1 a_n=\frac{1}{n^2+1} an=n2+11

∀ ϵ > 0 , ∣ a n − 0 ∣ = 1 n 2 + 1 < ϵ    ⟺    n > [ 1 ϵ − 1 ] , 即 lim ⁡ n → ∞ a n = 0 \forall \epsilon\gt 0,|a_n-0|=\frac{1}{n^2+1}\lt\epsilon\iff n\gt [\sqrt{\frac{1}{\epsilon}-1}],即\lim_{n\to\infin}a_n=0 ϵ>0,an0=n2+11<ϵn>[ϵ11 ],limnan=0

定义 { a n } \{a_n\} {an}数列,A为常数(constant),若 ∀ ϵ > 0 , ∃ N > 0 \forall \epsilon \gt 0, \exists N\gt 0 ϵ>0,N>0 n > N n\gt N n>N时, ∣ a n − A ∣ < ϵ |a_n-A|<\epsilon anA<ϵ

则称 A A A { a n } \{a_n\} {an}的极限,记作:
lim ⁡ n → ∞ a n = A \lim_{n\to\infin}a_n=A nliman=A

例1.
证明 lim ⁡ n → ∞ n 2 n + 1 = 1 2 \lim_{n\to\infin}\frac{n}{2n+1}=\frac{1}{2} limn2n+1n=21

∀ ϵ > 0 \forall\epsilon\gt 0 ϵ>0

∣ n 2 n + 1 − 1 2 ∣ = 1 2 ( 2 n + 1 ) < ϵ    ⟺    n > 1 2 ( 1 2 ϵ − 1 ) |\frac{n}{2n+1}-\frac{1}{2}|=\frac{1}{2(2n+1)}\lt\epsilon\iff n\gt \frac{1}{2}(\frac{1}{2\epsilon}-1) 2n+1n21=2(2n+1)1<ϵn>21(2ϵ11)
N = [ 1 2 ( 1 2 ϵ − 1 ) ] N=[\frac{1}{2}(\frac{1}{2\epsilon}-1)] N=[21(2ϵ11)],当 n > N n\gt N n>N
∣ n 2 n + 1 − 1 2 ∣ < ϵ |\frac{n}{2n+1}-\frac{1}{2}|\lt\epsilon 2n+1n21<ϵ

例2.
证明 l i m n → ∞ 2 n 2 − 1 2 n 2 + 1 = 1 lim_{n\to \infin}\frac{2n^2-1}{2n^2+1}=1 limn2n2+12n21=1

证: ∀ ϵ > 0 \forall\epsilon\gt0 ϵ>0

∣ 2 n 2 − 1 2 n 2 + 1 − 1 ∣ = 2 2 n 2 + 1 ≤ 1 n 2 < ϵ    ⟺    n > 1 ϵ |\frac{2n^2-1}{2n^2+1}-1|=\frac{2}{2n^2+1}\le\frac{1}{n^2}\lt\epsilon\iff n\gt\sqrt{\frac{1}{\epsilon}} 2n2+12n211=2n2+12n21<ϵn>ϵ1

N = [ 1 ϵ ] N=[\sqrt{\frac{1}{\epsilon}}] N=[ϵ1 ] n > N n\gt N n>N时,

∣ 2 n 2 − 1 2 n 2 + 1 − 1 ∣ < ϵ |\frac{2n^2-1}{2n^2+1}-1|<\epsilon 2n2+12n211<ϵ

∴ l i m n → ∞ 2 n 2 − 1 2 n 2 + 1 = 1 \therefore lim_{n\to \infin}\frac{2n^2-1}{2n^2+1}=1 limn2n2+12n21=1

4.1.2 数列极限的性质

极限是计算,凡计算就有性质,各种计算有自己的性质

  1. 唯一性

    l i m n → ∞ a n = A , l i m n → ∞ a n = B lim_{n\to\infin}{a_n}=A,lim_{n\to\infin}{a_n}=B limnan=A,limnan=B

    A = B A=B A=B

    证(反正法):设 A ≠ B A\ne B A=B,不妨设 A > B A\gt B A>B
    ϵ = A − B 2 > 0 \epsilon=\frac{A-B}{2}\gt 0 ϵ=2AB>0
    ∵ l i m n → ∞ a n = A \because lim_{n\to\infin}{a_n}=A limnan=A
    ∴ ∃ N 1 > 0 \therefore \exist N_1\gt 0 N1>0 n > N 1 n\gt N_1 n>N1
    ∣ a n − A ∣ < A − B 2    ⟺    A + B 2 < a n < 3 A − B 2 ( ∗ ) |a_n-A|<\frac{A-B}{2}\iff \frac{A+B}{2}<a_n<\frac{3A-B}{2}(*) anA<2AB2A+B<an<23AB()

    ∵ l i m n → ∞ a n = B \because lim_{n\to\infin}{a_n}=B limnan=B

    ∴ ∃ N 2 > 0 \therefore \exist N_2\gt 0 N2>0 n > N 2 n\gt N_2 n>N2

    ∣ a n − B ∣ < A − B 2    ⟺    3 B − A 2 < a n < A + B 2 ( ∗ ∗ ) |a_n-B|<\frac{A-B}{2}\iff\frac{3B-A}{2}\lt a_n\lt\frac{A+B}{2}(**) anB<2AB23BA<an<2A+B()

    N = m a x { N 1 , N 2 } N=max\{N_1,N_2\} N=max{N1,N2},当 n > N n\gt N n>N时, ( ∗ ) 、 ( ∗ ∗ ) (*)、(**) ()()均成立

    此时,矛盾, ∴ A = B \therefore A=B A=B

    例3: a n = ( − 1 ) n a_n=(-1)^n an=(1)n

    l i m n → ∞ a n lim_{n\to\infin}{a_n} limnan不存在

  2. 有界性,有极限则有界,有界不一定有极限

    l i m n → ∞ a n = A lim_{n\to\infin}{a_n}=A limnan=A,则 ∃ M > 0 \exist M\gt 0 M>0使 ∣ a n ∣ ≤ M |a_n|\le M anM,反之不成立

    证:取 ϵ = 1 \epsilon=1 ϵ=1

    ∵ l i m n → ∞ a n = A \because lim_{n\to\infin}{a_n}=A limnan=A

    ∴ ∃ N > 0 \therefore \exist N\gt 0 N>0,当 n > N n\gt N n>N ∣ a n − A ∣ < 1 ⇒ ∣ a n ∣ < 1 + ∣ A ∣ |a_n-A|<1\Rightarrow |a_n|\lt 1+|A| anA<1an<1+A

    M = m a x { ∣ a 1 ∣ , ∣ a 2 ∣ , … , ∣ a N ∣ , 1 + ∣ A ∣ } M=max\{|a_1|,|a_2|,\dots,|a_N|,1+|A|\} M=max{a1,a2,,aN,1+A}

    ∀ n \forall n n,有 a n ≤ M a_n\le M anM

    反例: a n = 1 + ( − 1 ) n a_n=1+(-1)^n an=1+(1)n l i m n → ∞ a n lim_{n\to\infin}{a_n} limnan不存在,但是 ∣ a n ∣ ≤ 2 |a_n|\le 2 an2

    三角不等式:
    ∣ ∣ a ∣ − ∣ b ∣ ∣ ≤ ∣ a ± b ∣ ≤ ∣ a ∣ + ∣ b ∣ \Big||a|-|b|\Big|\le\Big|a\pm b\Big|\le|a|+|b| aba±ba+b
    ∣ ∣ a n ∣ − ∣ A ∣ ∣ ≤ a n − A < 1 \Big||a_n|-|A|\Big|\le a_n-A\lt 1 anAanA<1
    ∣ a n ∣ − ∣ A ∣ < 1 |a_n|-|A|<1 anA<1

  3. 保号性

    l i m n → ∞ a n = A > 0 ( < 0 ) lim_{n\to\infin}{a_n}=A\gt 0(\lt 0) limnan=A>0(<0)

    则存在 ∃ N > 0 \exist N\gt 0 N>0,当 n > N n\gt N n>N时, a n > 0 ( < 0 ) a_n\gt 0(\lt 0) an>0(<0)

    证明:设 A > 0 A\gt 0 A>0

    ϵ = A 2 > 0 \epsilon=\frac{A}{2}\gt 0 ϵ=2A>0

    ∵ l i m n → ∞ a n = A \because lim_{n\to\infin}{a_n}=A limnan=A

    ∴ ∃ N > 0 \therefore\exist N\gt 0 N>0,当 n > N n\gt N n>N

    ∣ a n − A ∣ < A 2 ⇒ a n > A 2 > 0 |a_n-A|\lt\frac{A}{2}\Rightarrow a_n\gt\frac{A}{2}>0 anA<2Aan>2A>0

4.2 函数极限

4.2.1 定义

f ( x ) = x 2 − 1 x − 1 ( x ≠ 1 ) f(x)=\frac{x^2-1}{x-1}(x\ne 1) f(x)=x1x21(x=1)

f ( x ) = x + 1 ( x ≠ 1 ) f(x)=x+1(x\ne 1) f(x)=x+1(x=1)

d
x → 1 − x\to1^- x1, f ( x ) → 2 f(x)\to 2 f(x)2

x → 1 + ​ x\to1^+​ x1+, f ( x ) → 2 ​ f(x)\to 2​ f(x)2

x x x趋向于a的定义

x → { x ≠ a x → a − , x → a + ( 一 定 保 证 双 侧 趋 近 ) x\to\begin{cases} x\ne a \\ x\to a^-,x\to a^+(一定保证双侧趋近) \end{cases} x{x=axa,xa+()

证明: l i m x → 1 x 2 − 1 x − 1 = 2 lim_{x\to1}{\frac{x^2-1}{x-1}}=2 limx1x1x21=2

∀ ϵ > 0 \forall\epsilon\gt0 ϵ>0 ∣ x 2 − 1 x − 1 − 2 ∣ = ∣ x − 1 ∣ < ϵ |\frac{x^2-1}{x-1}-2|=|x-1|\lt\epsilon x1x212=x1<ϵ ( ϵ \epsilon ϵ为误差,具有任意性)

δ = ϵ \delta=\epsilon δ=ϵ,当 0 < ∣ x − 1 ∣ < δ 0\lt|x-1|\lt\delta 0<x1<δ时, ∣ x 2 − 1 x − 1 − 2 ∣ < ϵ |\frac{x^2-1}{x-1}-2|\lt\epsilon x1x212<ϵ,即 l i m x → 1 x 2 − 1 x − 1 = 2 lim_{x\to1}{\frac{x^2-1}{x-1}}=2 limx1x1x21=2

  1. 自变量趋于有限值时,函数的极限, x x x无限接近一个常数 a a a,函数无限接近的值

    y = f ( x ) ( x ∈ D ) y=f(x)(x\in D) y=f(x)(xD) A A A是常数

    ∀ ϵ > 0 \forall\epsilon\gt 0 ϵ>0 ∃ δ > 0 \exist \delta\gt 0 δ>0,当 0 < ∣ x − a ∣ < δ 0\lt|x-a|\lt\delta 0<xa<δ时,

    ∣ f ( x ) − A ∣ < ϵ |f(x)-A|\lt\epsilon f(x)A<ϵ

    A A A f ( x ) f(x) f(x) x → a x\to a xa时的极限

    记为: l i m x → a f ( x ) = A lim_{x\to a}{f(x)}=A limxaf(x)=A f ( x ) → A ( x → a ) f(x)\to A(x\to a) f(x)A(xa)

    注意:

    l i m x → a f ( x ) lim_{x\to a}{f(x)} limxaf(x) f ( x ) f(x) f(x) x = a x=a x=a处的定义无关

    如: l i m x → 2 x 3 − 8 x − 2 = l i m x → 2 ( x 2 + 2 x + 4 ) = 12 lim_{x\to 2}{\frac{x^3-8}{x-2}}=lim_{x\to 2}({x^2+2x+4})=12 limx2x2x38=limx2(x2+2x+4)=12

    左右极限:

    l i m x → a − f ( x ) ≜ f ( a − 0 ) lim_{x\to a^-}{f(x)}\triangleq f(a-0) limxaf(x)f(a0) - 左极限, x x x从左侧无限接近 a a a

    l i m x → a + f ( x ) ≜ f ( a + 0 ) lim_{x\to a^+}{f(x)}\triangleq f(a+0) limxa+f(x)f(a+0) - 右极限, x x x从右侧无限接近 a a a

    l i m x → a f ( x ) lim_{x\to a}{f(x)} limxaf(x)存在    ⟺    f ( a − 0 ) 、 f ( a + 0 ) \iff f(a-0)、f(a+0) f(a0)f(a+0)存在且相等,即极限存在的充要条件:左右极限都存在且二者相等

    例1. l i m x → 2 ( 3 x + 1 ) = 7 lim_{x\to 2}{(3x+1)}=7 limx2(3x+1)=7

    证: ∀ ϵ > 0 \forall\epsilon\gt 0 ϵ>0

    ∣ ( 3 x + 1 ) − 7 ∣ = 3 ∣ x − 2 ∣ < ϵ    ⟺    ∣ x − 2 ∣ < ϵ 3 |(3x+1)-7|=3|x-2|\lt\epsilon\iff |x-2|\lt\frac{\epsilon}{3} (3x+1)7=3x2<ϵx2<3ϵ

    δ = ϵ 3 \delta=\frac{\epsilon}{3} δ=3ϵ

    0 < ∣ x − 2 ∣ < δ 0\lt|x-2|\lt\delta 0<x2<δ时, ∣ ( 3 x + 1 ) − 7 ∣ < ϵ |(3x+1)-7|<\epsilon (3x+1)7<ϵ

    即: l i m x → 2 ( 3 x + 1 ) = 7 lim_{x\to2}{(3x+1)}=7 limx2(3x+1)=7

    例2. l i m x → 1 x 3 − 1 x − 1 = 3 lim_{x\to 1}\frac{x^3-1}{x-1}=3 limx1x1x31=3

    证: ∀ ϵ > 0 \forall\epsilon\gt0 ϵ>0

    ∣ x 3 − 1 x − 1 − 3 ∣ = ∣ x 2 + x + 1 − 3 ∣ = ∣ x 2 + x − 2 ∣ = ∣ x + 2 ∣ ∙ ∣ x − 1 ∣ < 4 ∣ x − 1 ∣ < ϵ    ⟺    ∣ x − 1 ∣ < ϵ 4 |\frac{x^3-1}{x-1}-3|=|x^2+x+1-3|=|x^2+x-2|=|x+2|\bullet|x-1|<4|x-1|\lt\epsilon\iff|x-1|\lt\frac{\epsilon}{4} x1x313=x2+x+13=x2+x2=x+2x1<4x1<ϵx1<4ϵ δ = m i n { ϵ 4 , 1 } \delta=min\{\frac{\epsilon}{4}, 1\} δ=min{4ϵ,1},当 0 < ∣ x − 1 ∣ < δ 0\lt|x-1|\lt\delta 0<x1<δ时, ∣ x 3 − 1 x − 1 − 3 ∣ < ϵ |\frac{x^3-1}{x-1}-3|<\epsilon x1x313<ϵ

    即: l i m x → 1 x 3 − 1 x − 1 = 3 lim_{x\to 1}\frac{x^3-1}{x-1}=3 limx1x1x31=3

  2. 自变量趋于无穷时,函数的极限
    x → ∞ { x → + ∞ x → − ∞ x → ± ∞ x\to\infty\begin{cases} x\to+\infty \\ x\to-\infty \\ x\to\pm\infty\end{cases} xx+xx±

    case 1. x → + ∞ x\to+\infty x+,当 ∀ ϵ > 0 \forall\epsilon\gt0 ϵ>0 ∃ X > 0 \exist X\gt0 X>0,当 x > X x\gt X x>X时, ∣ f ( x ) − A ∣ < ϵ |f(x)-A|\lt\epsilon f(x)A<ϵ,称 A A A f ( x ) f(x) f(x) x → + ∞ x\to+\infty x+时的极限

    例3 x → + ∞    ⟹    1 x → 0    ⟹    3 1 x → 1 x\to+\infty\implies\frac{1}{x}\to0\implies3^{\frac{1}{x}}\to1 x+x103x11

    例4 l i m x → + ∞ 2 x 2 x 2 + 1 = 2 lim_{x\to+\infty}{\frac{2x^2}{x^2+1}}=2 limx+x2+12x2=2

    证: ∀ ϵ > 0 \forall\epsilon\gt0 ϵ>0

    ∣ 2 x 2 x 2 + 1 − 2 ∣ = 2 x 2 + 1 < x 2 2 < ϵ    ⟺    x > 2 ϵ |\frac{2x^2}{x^2+1}-2|=\frac{2}{x^2+1}\lt\frac{x^2}{2}\lt\epsilon\iff x\gt\sqrt{\frac{2}{\epsilon}} x2+12x22=x2+12<2x2<ϵx>ϵ2

    X = 2 ϵ X=\sqrt{\frac{2}{\epsilon}} X=ϵ2 ,当 x > X x\gt X x>X时,

    ∣ 2 x 2 x 2 + 1 − 2 ∣ < ϵ |\frac{2x^2}{x^2+1}-2|\lt\epsilon x2+12x22<ϵ

    ∴ l i m x → + ∞ 2 x 2 x 2 + 1 = 2 \therefore lim_{x\to+\infty}{\frac{2x^2}{x^2+1}}=2 limx+x2+12x2=2

    case 2. x → − ∞ x\to -\infty x

    ∀ ϵ > 0 , ∃ X > 0 \forall \epsilon\gt0,\exist X\gt0 ϵ>0,X>0,当 x < − X x\lt -X x<X时,有 ∣ f ( x ) − A ∣ < ϵ |f(x)-A|<\epsilon f(x)A<ϵ,则称 A A A f ( x ) f(x) f(x) x → − ∞ x\to-\infty x时的极限,记作: l i m x → − ∞ f x = A lim_{x\to-\infty}{fx}=A limxfx=A

    case 3. x → ± ∞ ( ∞ )    ⟹    x 2 → + ∞    ⟹    e − x 2 → 0 x\to\pm\infty(\infty)\implies x^2\to+\infty\implies e^{-x^2}\to0 x±()x2+ex20

    ∀ ϵ > 0 , ∃ > 0 \forall\epsilon\gt0, \exist\gt0 ϵ>0,>0,当 ∣ x ∣ > X |x|\gt X x>X时,有 ∣ f ( x ) − A ∣ < ϵ |f(x)-A|<\epsilon f(x)A<ϵ,则称 A A A f ( x ) f(x) f(x) x → ∞ x\to\infty x时的极限,记作: l i m x → ∞ f x = A lim_{x\to\infty}{fx}=A limxfx=A

4.2.2 函数极限的性质

  1. 唯一性 - 函数极限存在即唯一

    l i m x → a f ( x ) = A , l i m x → a f ( x ) = B    ⟹    A = B lim_{x\to a}{f(x)}=A,\quad lim_{x\to a}{f(x)}=B\implies A=B limxaf(x)=A,limxaf(x)=BA=B

    l i m x → + ∞ f ( x ) = A , l i m x → + ∞ f ( x ) = B    ⟹    A = B lim_{x\to +\infty}{f(x)}=A,\quad lim_{x\to +\infty}{f(x)}=B\implies A=B limx+f(x)=A,limx+f(x)=BA=B

    l i m x → − ∞ f ( x ) = A , 1 m m l i m x → − ∞ f ( x ) = B    ⟹    A = B lim_{x\to -\infty}{f(x)}=A,\quad {1mm}lim_{x\to -\infty}{f(x)}=B\implies A=B limxf(x)=A,1mmlimxf(x)=BA=B

    l i m x → ∞ f ( x ) = A , l i m x → ∞ f ( x ) = B    ⟹    A = B lim_{x\to \infty}{f(x)}=A,\quad lim_{x\to \infty}{f(x)}=B\implies A=B limxf(x)=A,limxf(x)=BA=B

  2. 保号性

    l i m x → a f ( x ) > 0 ( < 0 )    ⟹    ∃ ϵ > 0 , 当 0 < ∣ x − a ∣ < ϵ 时 , f ( x ) > 0 ( < 0 ) lim_{x\to a}{f(x)}\gt0(\lt0)\implies \exist\epsilon\gt0,当0\lt|x-a|\lt\epsilon时,f(x)\gt0(\lt0) limxaf(x)>0(<0)ϵ>00<xa<ϵf(x)>0(<0)

    极限正,则去心领域正,极限负,则去心领域负

  3. 局部有界性

4.3 无穷小和无穷大

4.3.1 无穷小

l i m x → a α ( x ) = 0 lim_{x\to a}\alpha(x)=0 limxaα(x)=0,则 α ( x ) \alpha(x) α(x) x → a x\to a xa时为无穷小

l i m x → ∞ α ( x ) = 0 lim_{x\to\infty}\alpha(x)=0 limxα(x)=0,则 α ( x ) \alpha(x) α(x) x → ∞ x\to\infty x时为无穷小

3 ( x − 1 ) 2 3(x-1)^2 3(x1)2 l i m x → 1 3 ( x − 1 ) 2 = 0 lim_{x\to 1}{3(x-1)^2}=0 limx13(x1)2=0 3 ( x − 1 ) 2 3(x-1)^2 3(x1)2 x → 1 x\to 1 x1时为无穷小

0是无穷小, l i m 0 = 1 lim0=1 lim0=1

4.3.2 无穷大

l i m x → 1 f ( x ) = ∞ lim_{x\to 1}{f(x)}=\infty limx1f(x)=,则 f ( x ) f(x) f(x) x → a x\to a xa时为无穷大

l i m x → 1 1 f ( x ) = 0 lim_{x\to 1}{\frac{1}{f(x)}}=0 limx1f(x)1=0,则 f ( x ) f(x) f(x) x → a x\to a xa时为无穷大

无穷小和无穷大为倒数关系

4.4 极限运算法则

  1. l i m x → a / ∞ f ( x ) = A    ⟺    f ( x ) = A + α lim_{x\to a/\infty}{f(x)}=A\iff f(x)=A+\alpha limxa/f(x)=Af(x)=A+α,其中 α → 0 ( x → a / ∞ ) \alpha\to 0(x\to a/\infty) α0(xa/)

  2. α → 0 , β → 0 ,    ⟹    α ± β → 0 \alpha\to0,\beta\to0,\implies \alpha\pm\beta\to0 α0,β0,α±β0 x → a x\to a xa

  3. α → 0 ( x → a ) , ∣ β ∣ ≤ M    ⟹    α β → 0 ( x → a ) \alpha\to0(x\to a),|\beta|\le M\implies \alpha\beta\to0(x\to a) α0(xa),βMαβ0(xa)

  4. l i m f ( x ) = A , l i m g ( x ) = B lim{f(x)}=A,lim{g(x)}=B limf(x)=A,limg(x)=B,那么

    l i m [ f ( x ) ± g ( x ) ] = l i m f ( x ) ± l i m g ( x ) = A ± B lim{[f(x)\pm g(x)]}=lim{f(x)}\pm lim{g(x)}=A\pm B lim[f(x)±g(x)]=limf(x)±limg(x)=A±B

    l i m [ f ( x ) × g ( x ) ] = l i m f ( x ) × l i m g ( x ) = A × B lim{[f(x)\times g(x)]}=lim{f(x)}\times lim{g(x)}=A\times B lim[f(x)×g(x)]=limf(x)×limg(x)=A×B

    若又有 B ≠ 0 B\ne0 B=0,则

    l i m f ( x ) g ( x ) = l i m f ( x ) l i m g ( x ) = A B lim{\frac{f(x)}{g(x)}}=\frac{lim{f(x)}}{lim{g(x)}}=\frac{A}{B} limg(x)f(x)=limg(x)limf(x)=BA

  5. 如果 l i m f ( x ) lim{f(x)} limf(x)存在, c c c为常数,那么: l i m [ c f ( x ) ] = c l i m f ( x ) lim[{cf(x)}]=clim{f(x)} lim[cf(x)]=climf(x)

  6. 如果 l i m f ( x ) limf(x) limf(x)存在, n n n为正整数,那么: l i m [ f ( x ) ] n = [ l i m f ( x ) ] n lim[f(x)]^n=[limf(x)]^n lim[f(x)]n=[limf(x)]n

  7. φ ( x ) ≥ ψ ( x ) \varphi(x)\ge\psi(x) φ(x)ψ(x),而 l i m φ ( x ) = A , l i m ψ ( x ) = B lim{\varphi(x)}=A,lim{\psi(x)}=B limφ(x)=A,limψ(x)=B,那么 A ≥ B A\ge B AB

  8. (复合函数的极限运算法则)设函数 y = [ f ( x ) ] y=[f(x)] y=[f(x)]是由函数 u = g ( x ) u=g(x) u=g(x)与函数 y = f ( u ) y=f(u) y=f(u)复合而成, f [ g ( x ) ] f[g(x)] f[g(x)]在点 x 0 x_0 x0的某去心领域内有定义,若 l i m g ( x ) = u 0 lim{g(x)}=u_0 limg(x)=u0 l i m u → u 0 f ( u ) lim_{u\to u_0}{f(u)} limuu0f(u)=A,且存在 δ > 0 \delta\gt0 δ>0,当 x ∈ U ˚ ( x 0 , δ 0 ) x\in \mathring{U}(x_0,\delta_0) xU˚(x0,δ0)时,有 g ( x ) ≠ u 0 g(x)\ne u_0 g(x)=u0,则: l i m x → x 0 f [ g ( x ) ] = l i m u → u 0 f ( u ) = A lim_{x\to x_0}{f[g(x)]}=lim_{u\to u_0}{f(u)}=A limxx0f[g(x)]=limuu0f(u)=A

4.5 夹逼定理

数列

a n ≤ b n ≤ c n l i m n → ∞ a n = l i m n → ∞ c n = A }    ⟹    l i m n → ∞ b n = A \left. \begin{array}{r} a_n\le b_n\le c_n \\ lim_{n\to \infty}{a_n}=lim_{n\to\infty}{c_n}=A \end{array} \right \} \implies lim_{n\to\infty}{b_n}=A anbncnlimnan=limncn=A}limnbn=A

1 − 1 n ≤ 1 + 1 n ≤ 1 + 2 n 1-\frac{1}{n}\le 1+\frac{1}{n}\le 1+\frac{2}{n} 1n11+n11+n2

l i m n → ∞ ( 1 − 1 n ) = 1 l i m n → ∞ ( 1 + 2 n ) = 1 }    ⟹    l i m n → ∞ ( 1 + 1 n ) = 1 \left. \begin{array}{r} lim_{n\to\infty}{(1-\frac{1}{n})}=1 \\ lim_{n\to\infty}{(1+\frac{2}{n})}=1 \end{array} \right\} \implies lim_{n\to\infty}{(1+\frac{1}{n})}=1 limn(1n1)=1limn(1+n2)=1}limn(1+n1)=1

函数

{ f ( x ) ≤ g ( x ) ≤ h ( x ) l i m f ( x ) = l i m h ( x ) = A    ⟹    l i m g ( x ) = A \begin{cases} f(x)\le g(x) \le h(x) \\ lim{f(x)}=lim{h(x)}=A \end{cases} \implies lim{g(x)}=A {f(x)g(x)h(x)limf(x)=limh(x)=Alimg(x)=A

5. 导数

5.1 实例

  1. s = s ( t ) , t ∈ [ t 1 , t 2 ] s=s(t), t\in [t_1, t_2] s=s(t),t[t1,t2]

    t = t 0 t=t_0 t=t0,求 v ( t 0 ) v(t_0) v(t0)

    [ t 0 , t 0 + Δ ] [t_0,t_0+\Delta] [t0,t0+Δ] Δ s = s ( t 0 + Δ t ) − s ( t 0 ) \Delta s=s(t_0+\Delta t) - s(t_0) Δs=s(t0+Δt)s(t0)

    平均速度: v ˉ = Δ s Δ t \bar{v}=\frac{\Delta s}{\Delta t} vˉ=ΔtΔs

    瞬时速度: v ( t 0 ) = l i m Δ t → 0 Δ s Δ t v(t_0)=lim_{\Delta t\to 0}{\frac{\Delta s}{\Delta t}} v(t0)=limΔt0ΔtΔs

  2. y = f ( x ) y=f(x) y=f(x)

    k ˉ = f ( x 0 + Δ x ) − f ( x 0 ) Δ x \bar{k}=\frac{f(x_0+\Delta x)-f(x_0)}{\Delta x} kˉ=Δxf(x0+Δx)f(x0)

    Δ x → 0 \Delta x \to 0 Δx0时, k ˉ \bar{k} kˉ M 0 M_0 M0的切线,即:

    k 切 = l i m Δ x → 0 f ( x 0 + Δ x ) − f ( x 0 ) Δ x k_切=lim{\Delta x\to 0}{\frac{f(x_0+\Delta x)-f(x_0)}{\Delta x}} k=limΔx0Δxf(x0+Δx)f(x0)

    005

5.2 导数定义

y = f ( x ) ( x ∈ D ) y=f(x)(x\in D) y=f(x)(xD) x 0 ∈ D . x 0 + Δ x ∈ D x_0 \in D. x_0 +\Delta x \in D x0D.x0+ΔxD

Δ y = f ( x 0 + Δ x ) − f ( x 0 ) \Delta y=f(x_0 +\Delta x) -f(x_0) Δy=f(x0+Δx)f(x0)

l i m Δ x → 0 Δ y Δ x lim_{\Delta x\to0}{\frac{\Delta y}{\Delta x}} limΔx0ΔxΔy存在,称 f ( x ) f(x) f(x) x = x 0 x=x_0 x=x0处可导,表示切线斜率,表明变化率

极限值为 f ( x ) f(x) f(x) x = x 0 x=x_0 x=x0处的导数,记着: f ′ ( x 0 ) f'(x_0) f(x0) d y d x ∣ x = x 0 \frac{dy}{dx}|_{x=x_0} dxdyx=x0

注:

  1. f ′ ( x ) ≜ l i m x → x 0 f ( x ) − f ( x 0 ) x − x 0 f'(x)\triangleq lim_{x\to x_0}{\frac{f(x)-f(x_0)}{x-x_0}} f(x)limxx0xx0f(x)f(x0)

  2. f ′ ( x 0 ) f'(x_0) f(x0)存在,则 f ( x ) f(x) f(x) x = x 0 x=x_0 x=x0处连续(在 x 0 x_0 x0处,极限值=函数值),可导一定连续,连续不一定可导,即:可导是连续的充分非必要条件

  3. 左导数: f _ ′ ( x 0 ) = l i m Δ x → 0 − Δ y Δ x f'_\_(x_0)=lim_{\Delta x\to 0^-}{\frac{\Delta y}{\Delta x}} f_(x0)=limΔx0ΔxΔy

  4. 左导数: f + ′ ( x 0 ) = l i m Δ x → 0 + Δ y Δ x f'_+(x_0)=lim_{\Delta x\to 0^+}{\frac{\Delta y}{\Delta x}} f+(x0)=limΔx0+ΔxΔy

  5. f ′ ( x 0 ) f'(x_0) f(x0)存在    ⟺    \iff f _ ′ ( x 0 ) f'_\_(x_0) f_(x0) f + ′ ( x 0 ) f'_+(x_0) f+(x0)存在且相等(左右导数)

    左右导数不等的例子

    f ( x ) = { e x − 1 , x < 0 l n ( 1 + 2 x ) , x ≥ 0 f(x)=\left \{ \begin{array}{l} e^x - 1,x\lt 0 \\ ln{(1+2x)}, x\ge 0 \end{array} \right. f(x)={ex1,x<0ln(1+2x),x0

    f _ ′ ( 0 ) = l i m x → 0 − f ( x ) − f ( 0 ) x − 0 = 1 f'_\_(0)=lim_{x\to 0^-}{\frac{f(x)-f(0)}{x-0}}=1 f_(0)=limx0x0f(x)f(0)=1

    f + ′ ( 0 ) = l i m x → 0 + f ( x ) − f ( 0 ) x − 0 = 2 f'_+(0)=lim_{x\to 0^+}{\frac{f(x)-f(0)}{x-0}}=2 f+(0)=limx0+x0f(x)f(0)=2

    左右导数都存在但不相等,所以导数 f ′ ( 0 ) f'(0) f(0)不存在

  6. 瞬时速度: v ( t 0 ) = s ′ ( t 0 ) v(t_0)=s'(t_0) v(t0)=s(t0)

    瞬时加速度: a ( t 0 ) = v ′ ( t 0 ) a(t_0)=v'(t_0) a(t0)=v(t0)

    f ( x ) f(x) f(x) M 0 ( x 0 , f ( x 0 ) ) M_0(x_0,f(x_0)) M0(x0,f(x0))的切线: y − f ( 0 ) = f ′ ( x 0 ) ( x − x 0 ) y-f(0)=f'(x_0)(x-x_0) yf(0)=f(x0)(xx0)

5.3 求导数

5.3.1 定义求导数

根据定义计算对应的极限值即可

y = f ( x ) = x n y=f(x)=x^n y=f(x)=xn

f ′ ( x ) = n x n − 1 f'(x)=nx^{n-1} f(x)=nxn1

x n − a n = ( x − a ) ( x n − 1 + a x n − 2 + ⋯ + a n − 1 ) x^n-a^n=(x-a)(x^{n-1}+ax^{n-2}+\dots+a^{n-1}) xnan=(xa)(xn1+axn2++an1)

例: x 5 − a 5 = ( x − a ) ( x 4 + a x 3 + a 2 x 2 + a 3 x + a 4 ) x^5-a^5=(x-a)(x^4+ax^3+a^2x^2+a^3x+a^4) x5a5=(xa)(x4+ax3+a2x2+a3x+a4)

5.3.2 求导法则

初等函数求导:

幂函数: ( x a ) ′ = a n a − 1 , a ∈ R (x^a)'=an^{a-1},a\in R (xa)=ana1,aR,特殊: ( x ) ′ = 1 2 x (\sqrt{x})'=\frac{1}{2\sqrt{x}} (x )=2x 1 ( 1 x ) ′ = 1 x 2 (\frac{1}{x})'=\frac{1}{x^2} (x1)=x21

指数函数: ( a x ) ′ = a x l n ( a ) (a^x)'=a^xln(a) (ax)=axln(a),特殊: ( e x ) ′ = e x (e^x)'=e^x (ex)=ex

对数函数: ( l o g a x ) ′ = 1 x l n ( a ) , ( a > 0 且 a ≠ 1 ) (log_{a}{x})'=\frac{1}{xln(a)},(a>0且a\ne 1) (logax)=xln(a)1,(a>0a=1),特殊: ( l n ( x ) ) ′ = 1 x (ln(x))'=\frac{1}{x} (ln(x))=x1

三角函数:

( s i n ( x ) ) ′ = c o s ( x ) (sin(x))'=cos(x) (sin(x))=cos(x)

反三角函数:

6. 微分

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值