Calculus 微积分

Calculus 微积分

微积分包含微分(differentiation)和积分(integration),微分是基于变量增量无限细微化思想来分析解决问题的方法,积分是基于细微化增量无限汇聚思想来分析解决问题的方法。

在微分理论中,一个变量的无限小的增量 \(\Delta x\) 被称为“微分”(Differentiation),一般记为 \(\mathrm{d} x\)

Limit 极限

The limit of a function 函数极限

[Def] limit: Given a function, \(f(x):\R\mapsto \R\) ,

\[\lim_{x\to a}f(x)=L \iff \forall \epsilon >0 \exists \delta >0: 0<|x-a|<\delta \implies |f(x)-L|<\epsilon \]

\[\lim_{x\to x_0} f(x)= A\in \R\cup\{+\infty,-\infty\} \iff \lim_{x\to x_0^+} f(x)=A \wedge \lim_{x\to x_0^-} f(x)=A \]

[Def] limit for multivariate functions 多元函数极限:
Given a multivariate function \(f(\bm x):\R^n\mapsto \R\) , and its domain \(D\) :

\[\lim_{\bm x \to \bm a} f(\bm x)=L \iff \forall \epsilon \exists\delta: 0<\|\bm x-\bm a\|_2=\sqrt{\sum_{i}^n (x_i-a_i})<\delta \implies |f(\bm x)-L|<\epsilon \]

. 也就说,用来限制“邻域”范围的是欧氏距离。

One-side limit 单侧极限
极限运算性质

假设 \(\lim_{x\to a} f(x), \lim_{x\to a} g(x)\) 均存在,则

\[\begin{aligned} &\lim _{x \rightarrow a}[f(x)+g(x)]=\lim _{x \rightarrow a} f(x)+\lim _{x \rightarrow a} g(x) \\ &\lim _{x \rightarrow a}[f(x)-g(x)]=\lim _{x \rightarrow a} f(x)-\lim _{x \rightarrow a} g(x) \\ &\lim _{x \rightarrow a}[c f(x)]=c \lim _{x \rightarrow a} f(x) \\ &\lim _{x \rightarrow a}[f(x) g(x)]=\lim _{x \rightarrow a} f(x) \cdot \lim _{x \rightarrow a} g(x) \\ &\lim _{x \rightarrow a} \frac{f(x)}{g(x)}=\frac{\lim _{x \rightarrow a} f(x)}{\lim _{x \rightarrow a} g(x)} \text { if } \lim _{x \rightarrow a} g(x) \neq 0 \end{aligned} \]

。运用上述公式时一定要记得先验证先决条件的成立性。

The Squeee Theorem 夹逼定理

若在a的去心邻域内处处有 \(f(x)\le g(x)\le h(x)\) ,且 \(\lim_{x\to a}f(x)=\lim_{x\to a}h(x)=L\) ,则 \(\lim_{x\to a} g(x)=L\) .

洛必达法则

洛必达法则(L'Hospital's Rule)以法国数学家洛必达命名,但实际上是由瑞士数学家伯努利(Bernoulli)发现的。

法则内容:记扩展实数集 \(\bar\R=\R\cup\{\infty, -\infty\}\) , 设 \(c\in \bar\R\) , \(f(x), g(x)\)\(x=c\) 附近可微,其导数记为 \(f'(x),g'(x)\) ,且有 \(\lim_{x\to c}\frac{f'(x)}{g'(x)}\in \bar\R, g'(x)\ne 0\) 。如果 \(\lim_{x\to c}{f(x)}=\lim_{x\to c}g(x)=0\)\(\lim_{x\to c}|f(x)|=\lim_{x\to c}|g(x)|=\infty\) (换句话说, \(\frac{f(x)}{g(x)}\)\(\frac00\) 型或 \(\frac{\infty}{\infty}\) 型),则有

\[\lim_{x\to c}\frac{f(x)}{g(x)}=\lim_{x\to c} \frac{f'(x)}{g'(x)} \]

\(0\cdot\infty, \infty-\infty, 0^0, 1^{\infty}, \infty^0\) 型的未定式均可转换为一般型 \(\frac{0}{0}, \frac{\infty}{\infty}\) 求解。参见 https://en.wikipedia.org/wiki/L'Hôpital's_rule 中的 "Other indeterminate forms" 部分。

重要极限

\[\lim_{x\to 0} \frac{\sin x}{x} = 1 \\ \]

留意函数 \(y=\sin x /x (x\neq 0)\) 曲线形状尤其在x=0附近形状与极限值 \(\lim_{x\to 0}\sin x/x\) 的关系。

高阶无穷小、低阶无穷小、同阶无穷小、等价无穷小

若两个去穷小量 $ f(x), g(x) $ 满足

\[\lim_{f(x)\to 0, g(x)\to 0}\frac{f(x)}{g(x)}= 0 \]

,则称f(x)是g(x)的高阶无穷小量,g(x)是f(x)的低阶无穷小量,即在他们逐步趋近于0时,高阶无穷小量比低阶无穷小量趋近得更快。

若 $\lim_{f(x)\to 0, g(x)\to 0}\frac{f(x)}{g(x)}= 1 $ ,则称f(x)和g(x)是等价无穷小量。
若 $\lim_{f(x)\to 0, g(x)\to 0}\frac{f(x)}{g(x)}= c (c>0) $ ,则称f(x)和g(x)是同阶无穷小量。
若 $\lim_{f(x)\to 0, g(x)\to c (c>0)}\frac{f(x)}{[g(x)]^k}= c (c>0) $ ,则称f(x)是g(x)的k阶无穷小量。

Techniques of Evaluating Limits 求解极限的技巧
  1. 直接代入。
  2. 根式有理化。
  3. 因式分解。
  4. 化解为常见重要标准形式。
  5. 洛必达法则。
  6. 函数有界+运算性质+夹逼定理。

连续性 continuity
[Def] 称f(x)在x=a处连续,如果

\[\lim_{x\to a} = f(a) \]

。多元变量函数亦如此。

称f(x)在集合或区间上连续,当其在集合或区间内每点处都连续。

[Theorem] 如果函数f是连续的,则其逆函数(如果存在)也是连续的。

连续不蕴含可导,如y=|x|在0处连续但不可导。

[Theorem]

\[\lim_{x\to a} f(g(x)) = f(\lim_{x\to a} g(x)), \text{ if } f \text{ is continuous at } \lim_{x\to a}g(x) \]

渐近线 asymptote

水平渐近线
如果

\[\lim_{x\to +\infty} f(x)=c \text { 或 } \lim_{x\to -\infty} f(x)=c \]

,则称水平线(x轴平行线) \(y=c\) 是函数 \(f(x)\) 的水平渐近线。

竖直渐近线
如果

\[\lim_{x\to a} f(x)=+\infty \text { 或 } \lim_{x\to a} f(x)= -\infty \]

,则称竖直线(y轴平行线) \(x=c\) 是函数 \(f(x)\) 的竖直渐近线。

(用图直观展示)

Derivative 导数

导数被创建来研究变量的变化率的工具。

[Def] Derivative: For a function \(f: \R\mapsto\R, x\mapsto f(x), x\in \R\) , the derivative of \(f\) with respect to \(x\) is defined as the limit

\[\frac{\mathrm{d}f}{\mathrm{d} x} := \lim_{\Delta x\to 0} \frac{f(x+\Delta x)-f(x)}{\Delta x} \]

.

In calculus, we also denote the derivative of a function \(f(x)\) w.r.t. \(x\) simply as \(f'(x)\) .

求解单变元函数的导数的工具:导数相关数学工具;求解以方程刻画的两个变元间的纠缠关系(如 \(x^2+y^2=4\) )的导数:微分方程。

[Def] Partial Derivative: For a real-valued function of multivariate \(f: \R^n \mapsto \R, \bm x \mapsto f(\bm x), \bm x = (x_1,x_2,..., x_n)\) , the partial derivative of \(f\) with respect to \(x_i\) is defined as

\[\frac{\partial f}{\partial x_i}= \lim_{\Delta\to 0} \frac{f(x_1, ...,x_{i-1}, x_i +\Delta, x_{i+1}, ...,x_n) - f(x_1, ...,x_{i-1}, x_i , x_{i+1}, ...,x_n)}{\Delta} \]

The gradient of a real-valued function $f: \R^n \mapsto \R, \bm x\mapsto f(\bm x), \bm x\in \R^{n\times 1}=[x_1, x_2,...,x_n]^T $ with respect to the colunn vector \(\bm x\) is defined as a row vector of partial derivatives:

\[\nabla_{\bm x} f = \frac{\mathrm{d} f}{\mathrm{d} \bm x}:=\left[ \frac{\partial f}{\partial x_1}, ..., \frac{\partial f}{\partial x_n}\right] \in \R^{1\times n} \]

.

注意,梯度是各分量偏导数的向量形式。

It is not uncommon to define the gradient vector as a column vector, following the convension that vectors are generally column vectors.

The reason why we define the gradient vector as a row vector is twofold: (1) First, we can consistently generalize the gradient to vector-valued functions \(f:\R^n \mapsto \R^m\) (then
the gradient becomes a matrix). (2) Second, we can immediately apply the multi-variate chain rule without paying attention to the dimension of the gradient.

标量对列向量的梯度向量被定义为行向量形状。
标量对列向量的导数也常被定义为列向量,但定义为行向量时有两个优点,(1)能一致地将标量对列向量的梯度(向量)推广到列向量对列向量梯度(此时为矩阵);(2)应用链式法则时无需费心于梯度矩阵与向量乘法上的维度适配。

Hessian: For a real-valued function \(f: \R^n \mapsto \R, \bm x\mapsto f(\bm x)\) , the Hessian is defined as the second-derivative of \(f\) with respect to \(\bm x\)

\[H\in\R^{n\times n} \text{ , where } H_{i,j}=\frac{\partial^2 f}{\partial x_i \partial x_j} \]

is a symmetric matrix.

If \(f: \R^{n } \mapsto \R^{m }, \bm x \mapsto f(\bm x)\) , then Jacobian

\[J \in\R^{m \times n} \text{ , where } J_{i,j}=\frac{\partial f_i}{\partial x_j} \]

and the Hessian

\[H\in\R^{m\times n\times n} \text{ , where } H_{ijk}=\frac{\partial^2 f_i}{\partial x_j \partial x_k} \]

is an ( \(m\times n\times n\) )-tensor.

In the context of neural networks, where the input dimensionality is often much higher than the dimensionality of the labels, the reverse mode is computationally significantly cheaper than the forward mode.

机器学习中的后向传播算法是数值分析中“自动微分”的一个特例。

[Theorem] 多变元复合函数的链式法则:
For a real-valued funtion \(f(\bm x ): \R^n\mapsto \R, \bm x=[x_1,x_2,\dots,x_n]\) , and \(\bm x(\bm u):\R^m\mapsto \R^n=[ x_1(\bm u), x_2(\bm u), \dots, x_n(\bm u)], \bm u=[u_1,\dots,u_m]\) , then

\[\frac{\partial{f}}{\partial{u_j}}=\sum_i^n \frac{\partial{f}}{\partial{x_i}}\frac{\partial{x_i}}{\partial{u_j}} \\ \frac{\partial{f}}{\partial{\bm{\vec{u}}}}=\frac{\partial{f}}{\partial{\bm{\vec{x}}}}\frac{\partial{\bm{\vec{x}}}}{\partial{\bm{\vec{u}}}} \]

, where \(\frac{\partial{f}}{\partial{\bm{\vec{u}}}}\in\R^{1\times m}\) is a row-shaped vector of gradient of \(f\) w.r.t. \(\bm{\vec u}\) , and \(\frac{\partial{\bm{\vec{x}}}}{\partial{\bm{\vec{u}}}}\in\R^{m\times n}\) is the Jacobian matrix of \(\bm{\vec x}\) w.r.t. \(\bm{\vec u}\) .

total differential: for a real-valued function of independent multivariables, \(f(\bm x)\) ,

\[df=\nabla_{\bm x} f\cdot \bm{dx}=\sum_i \frac{\partial f}{\partial x_i}dx_i \]

.

[Def] 多变元函数的方向导数 directional derivative: For a function \(f(\bm{\vec x}):\R^n\mapsto \R\) , and a constant unit vector (direction) \(\bm{\vec u}\) , the directional derivative of \(f(\bm{\vec x})\) in the direction of a unit vector \(\bm{\vec u}\) is

\[D_{\bm u}f(\bm x):=\lim_{h\to 0} \frac{f(\bm x+h\bm u)-f(\bm x}{h} \]

.

And the partial derivatives come to be a special case of directional derivatives. (let \(\bm u=[0,\dots, 0,1,0,\dots,0])\) .

[Theorem] For \(f(\bm x)\) and a unit vector \(\bm u\) , the directional derivative

\[D_{\bm u}f(\bm x)=\nabla_{\bm x} f \cdot\bm u \]

级数 Series: 无穷个数项的和:

\[\sum_{n}^{\infty} a_n :=\lim_{N\to \infty} \sum_n^N a_n \]

自然常数e Natural Nubmer e

\[e = \lim_{n\to \infty}{(1+\frac1{n})^n}=\sum_{n=0}^{\infty}\frac1{n!} \]

几何级数 geometric series:

\[\sum_{n=0}^{\infty}ar^n=a+ar+ar^2+\cdots \]

is convergent if \(|r|<1\) , and the sum is

\[\sum_{n=0}^{\infty}ar^n=\frac{a}{1-r} ,\; |r|<1 \]

. If \(|r|\ge 1\) , the geometric series is divergennt.

Talor series 泰勒级数:
For a real-valued function \(f: \R \mapsto \R, x\mapsto f(x)\) , the Taylor series at \(x_0\) is defined as

\[f(x)=\sum_{k=0}^\infty \frac{D_x^k f(x_0)}{k!}(x-x_0)^k \]

, where \(D_x^k f(x_0)\) is the k-th derivative of f with respect to \(x\) , evaluated at \(x_0\) .

Multivariant Talor series: For a real-valued function \(f: \R^n \mapsto \R, \bm x\mapsto f(\bm x), \bm x\in\R^n\) , the multivariant Talor series at \(\bm x_0\in\R^n\) is defined as

\[f(\bm x)=\sum_{k=0}^\infty \frac{D_{ \bm x}^k f(\bm x_0)}{k!} (\bm x- \bm x_0)^{\otimes k} \]

, where \(D_{\bm x}^k f(\bm x_0)\) is the k-th (total) derivative of f with respect to \(\bm x\) , evaluated at \(\bm x_0\) . \((\bm x-\bm x_0)^{\otimes k}\) is the result of applying outer product on \(k\) times of the vector \((\bm x -\bm x_0)\) . Note that \(D_x^k f(\bm x)\) and \((\bm x-\bm x_0)^{\otimes k}\) are both k-th order tensor, i.e. \(D_x^k f(\bm x), (\bm x-\bm x_0)^{\otimes k} \in \R^{\overbrace{n\times ...\times n}^{k \text{ times}}}\) .

\[\left[D_{\bm x}^k f(\bm x_0)\right] \left[(\bm x- \bm x_0)^{\otimes k}\right] = \sum_{i_1=1}^n...\sum_{i_k=1}^n \left[D_{\bm x}^k f(\bm x_0) \right]_{i_1,...,i_k} \left[(\bm x-\bm x_0)^{\otimes k}\right]_{i_1,...,i_k} \]

For a vector $ \bm x \in \R^n$ , the outer product is defined as

\[\bm x^{\otimes k} = \overbrace{\bm x\otimes ...\otimes \bm x}^{k \text{ times}}=\mathbf Y \in \R^{\overbrace{n\times ...\times n}^{k \text{ times}}}, \text{ where } Y_{i_1,...,i_k}=x_{i_1}x_{i_2}...x_{i_k} \]

.

Derivative Properties 导数运算性质

\[\{a\}' = 0 \\ \{e^x\}'=e^x \\ \{a^x\}' = a^x \ln x \\ \{x^n\}' = n x^{n-1}, n\neq 0 \wedge n\in \Z \\ \{a f(x)\}'=a f'(x) \\ \{f(x)\pm g(x)\}' = f'(x) \pm g'(x) \\ \{f(g(\cdot))\}' = f'(\cdot) g'(\cdot) \\ \{f(x)g(x)\}' = f'(x)g(x)+f(x)g'(x) \\ \{\frac{f(x)}{g(x)} \}'=\{f(x)g^{-1}(x)\}' = \frac{1}{g^2(x)}\left[f'g-fg'\right] \]

Differentiation Properties 微分运算性质

[Theorem] For a differentiable function \(x\mapsto f(x)\) , we have

\[\mathrm{d} f(x) = \frac{\mathrm{d} f(x)}{\mathrm{d} x} \cdot \mathrm{d} x = f'(x) \mathrm{d} x \]

.

结合导数运算性质可得微分运算性质(以下公式中的字母d一般表示微分符号):

\[d[e^x] = [e^x] dx \\ d[a^x] = [a^x \ln a] dx \\ \]

临界点 critical point: 一阶导数为0或不存在的点,分一阶导数为0的点称驻点(stationary point),导数不存在的点称奇点(singular point)。
驻点 stionary point:在该点处一阶导数存在且为0。
(争议:有将驻点视为临界点的说法。)
奇点 singular point: (对于一元函数)导数不存在的点。
拐点 inflection point: 二阶导数在该点两边异号。
极值 extremum(局部最值 local maximum/minimum): 极大值/极小值点:在点邻域内该点函数值最大/最小。
最值(全局最值 global maximum/minimum):最大值/最小值点:在函数整个定义域内该点函数值最大/最小。(全局)最大值或最小值也可能有多个,如函数sin x的最值。

plural forms: extrema, maxima, minima.

极值点处的一阶导数(如果存在)为0,但一阶导数为0的点不一定是极值点。对于凸函数或凹函数,一阶导数(如果存在)为0的点等价于极值点。

Lagragian Mean Value Theorem 拉格朗日中值定理

函数若在闭区间[a,b]上连续且在其去端开区间(a,b)上可导,则在(a,b)间存在一个值c使得函数在该点处的切线的斜率与割线的斜率相同。

弧微分 arc differential: 思想:以线段近似弧长。

\[(ds)^2=(dx)^2+(dy)^2 \\ ds=\sqrt{(dx)^2+\left(\frac{dy}{dx}dx\right)^2}=\left(\sqrt{1+\left(\frac{dy}{dx}\right)^2}\right)dx \]


(figure source: Baidu-Baike)

Integration 积分

\[\int_a^b f(x)dx := \lim_{n\to \infty} \sum_i^n f(x_i^*)\Delta x, \Delta x=\frac{b-a}{n} \]

Definite Integration Properties 定积分性质

\[\int_a^b f(x)dx = \lim_{\epsilon\to 0} \int_{a+\epsilon}^b f(x)dx \\ \int_b^a f(x)dx = -\int_a^b f(x)dx \]

Indefinite Integration Properties 不定积分运算性质

不定积分等式 \(\int f(x)dx = F(x) +C\) 的含义实际上是指,以 \(F(x)\) 中的变量x作为积分上界对 \(f(x)\) 积分,因其没有指定下界导致一个常数C的出现,换言之, \(F(x)+C=\int_?^x f(t)dt=\int f(x)dx\)

\[ \left[\int f(x) d x\right]^{\prime}=f(x) \\ d \int f(x) d x=f(x) d x \\ \int F^{\prime}(x) d x=F(x)+C \\ \int d F(x)=F(x)+C \\ \int f(g(x))g'(x)dx = \int f(u)du \]

[Theorem] The Substitution Rule.

If \(u=g(x)\) is differentiable function with range being an interval and \(f\) is continuous on the interval, then $\int f(g(x))g'(x)dx = \int f(u)du $ .

常见函数不定积分公式

\[\int \frac{d F(x)}{d x} d x=F(x)+C \\ \int a d x=a x+c \\ \int x^{n} d x=\frac{1}{n+1} x^{n+1}+C, n \neq-1 \wedge n \in \Z \\ \int x^{k} d x=\frac{1}{k+1} x^{k+1}+C, k \neq-1 \wedge k \in \mathbb{Q} \\ \int \frac{1}{x} d x=\ln |x|+C \\ \int e^{x} d x=e^{x}+C \\ \int a^{x} d x=\frac{a^{x}}{\ln a}+C, a>0 \wedge a \neq 1 \\ \int \sin x d x=-\cos x +C \\ \int \cos x d x=\sin x+C \]

常见凑微分公式

\[\int f(a x+b) d x=\frac{1}{a} \int f(a x+b) d(a x+b)(a \neq 0) \\ \int f\left(a x^{2}+b\right) x d x=\frac{1}{2 a} \int f\left(a x^{2}+b\right) d\left(a x^{2}+b\right)(a \neq 0) \\ \int f\left(a x^{n}+b\right) x^{n-1} d x=\frac{1}{n a} \int f\left(a x^{n}+b\right) d\left(a x^{n}+b\right)(a\neq 0, n \neq 0) \\ \int f\left(\frac{1}{x}\right) \frac{1}{x^{2}} d x=-\int f\left(\frac{1}{x}\right) d\left(\frac{1}{x}\right) \\ \int f(\sqrt{x}) \frac{1}{\sqrt{x}} d x=2 \int f(\sqrt{x}) d(\sqrt{x}) \\ \int f(\ln x) \frac{1}{x} d x=\int f(\ln x) d(\ln x) \\ \int f\left(e^{a x}\right) e^{a x} d x=\frac{1}{a} \int f\left(e^{a x}\right) d\left(e^{a x}\right)(a \neq 0) \\ \int f(\sin x) \cos x d x=\int f(\sin x) d(\sin x) \\ \int f(\cos x) \sin x d x=-\int f(\cos x) d(\cos x) \]

体积 Volume

Cylindric shell method 旋转体体积求解

\[V=\int_a^b \underbrace{f(x)}_{周长circumference} \underbrace{h(x)}_{高 height} \underbrace{dx}_{厚 thickness} \]

比较适合用于经旋转形成的各类复杂旋转体(cylindric solid)。

旋转体侧面积:被旋转曲线所旋转形成的闭曲面的面积。

侧面积的计算中用的微元是弧微分ds,而体积计算中用的微元是自变量微分dx。

Differential Equations 微分方程

斜率场 Direction Field (Slope Field) 方法:y(x)是需求解的未知量,导数 \(dy/dx\) 是关于x、y的已知函数,即已知 \(\frac{dy}{dx}=F(x,y)\) ,则可以x,y为笛卡尔坐标系,采样若干(x,y)数据点,计算相应的 \(dy/dx\) ,对于每个数据点(x,y),在坐标点(x,y)画出以该点处导数 \(dy/dx|_{(x,y)}\) 为斜率的短线段,可从整个图像中根据导数走向看出(逼近出)未知量y(x)的大致方向(可能有多个解的曲线)。
例: https://www.geogebra.org/m/W7dAdgqc

变元(自变及因变)可拆分到等式两端的微分方程可通过将变元拆分到等式两端后对等式两端做积分而求得解析解,如

\[\frac{dy}{dx}=f(x) g(y) \iff h(y)dy=f(x)dx \text{ 若 } g(y)\ne 0 \\ \implies \int h(y)dy = \int f(x)dx \]

正交轨线 orthogonal trajectory: 一条曲线总正交地相交于一簇曲线中的每一条,则称这条曲线是这一簇曲线的正交轨线(可能有多条,也称正交轨线簇)。

两条直线正交则它们的斜率乘积为-1。

求解正交轨线的过程:对于给定一簇曲线 F(x,y,c)=0,其中x,y是自变量和因变量,c是定义簇的任意常量(相对于x,y来说是常量),求解出关于项 \(\frac{dy}{dx}\) 的微分方程 \(H(\frac{dy}{dx}, x, c)=0\) ,然后将其中的项 \(\frac{dy}{dx}\) 换为 \(-\frac{dx}{dy}\) 得到另一个微分方程 \(H(-\frac{dx}{dy},x,y,c)=0\) ,并且联立F(x,y,c)=0(因为是交点,则对于交点(x,y)其既在正交轨线上也在曲线F(x,y,c)=0上),消除c后解出该微分方程即是曲线F(x,y,c)=0的正交轨线。

例:求圆曲线 \(F(x,y,r)=0:x^2+y^2=r^2\) 的正交轨线簇。
解:方程两边取微分 \(\frac{d\cdot}{dx}\)\(2x+2y\frac{dy}{dx}=0\) ,该方程即是导数(切线斜率)方程,而正交轨线意味着相交时正交,则变换导数形成关于正交轨线的微分方程 \(2x-2y\frac{dx}{dy}=0\) ,求解该方程(定义簇的圆半径r本没有出现故无需再联立 \(x^2+y^2=r^2\) ),

\[2x-2y\frac{dx}{dy}=0\implies \frac{dx}{x}=\frac{dy}{y}\implies \\ \int\frac{dx}{x}=\int\frac{dy}{y}\implies \ln|y|+C_1=\ln|x|+C_2\implies |y|=e^{C_3}|x| \implies y=Cx \]

,即正圆簇 \(x^2+y^2=r^2\) 的正交轨线是曲线簇 \(y=Cx\)

从正圆例子中还看出,正圆的导数方程(微分方程)没有出现圆半径r,即不同半径的正圆的导数曲线是相同的。

一阶线性微分方程 first-order Linear differential equation

the form

\[\frac{dy}{dx}+P(x)y=Q(x) \]


书籍特点

《Calculus》 by James Stewart, 以丰富的示例进行讲解,其定义概念的风格大多先是描述性、直观性思维方式,再者形式化语言描述,对入门者较友好。


References:

  • Calculus, by James Stewart, 8th edition.
posted @ 2022-07-06 23:41  二球悬铃木  阅读(400)  评论(0编辑  收藏  举报