极值

f(x)在U(x0)有定义,∀x∈U˚f(x)\text{在}U(x_0)\text{有定义},\forall x\in\mathring{U}

若f(x)<f(x0)⇒f(x0)是极大值,x=x0是极大点\text{若}f(x)<f(x_0)\Rightarrow f(x_0)\text{是极大值},x=x_0\text{是极大点}

若f(x)>f(x0)⇒f(x0)是极小值,x=x0是极小点\text{若}f(x)>f(x_0)\Rightarrow f(x_0)\text{是极小值},x=x_0\text{是极小点}

定理一

f(x)在x0处可导,若f(x)在x0处取得极值⇒f′(x0)=0f(x)\text{在}x_0\text{处可导,若}f(x)\text{在}x_0\text{处取得极值}\Rightarrow f'(x_0)=0

f′(x0)=0⇏f(x)在x0处取得极值f'(x_0)=0\nRightarrow f(x)\text{在}x_0\text{处取得极值}

证明方法:费马引理,请读者自行验证

定理二

设f(x)在x0处连续,且U˚(x0,δ)可导\text{设}f(x)\text{在}x_0\text{处连续,且}\mathring{U}(x_0,\delta)\text{可导}

x∈(x0−δ,x0),f′(x)>0x∈(x0,x0+δ),f′(x)<0}⇒f(x)在x=x0处取得极大值\left. \begin{matrix} x \in (x_0-\delta,x_0),f'(x)>0 \\ x \in (x_0,x_0+\delta),f'(x)<0 \\ \end{matrix} \right\}\Rightarrow f(x)\text{在}x=x_0\text{处取得极大值}

x∈(x0−δ,x0),f′(x)<0x∈(x0,x0+δ),f′(x)>0}⇒f(x)在x=x0处取得极小值\left. \begin{matrix} x \in (x_0-\delta,x_0),f'(x)<0 \\ x \in (x_0,x_0+\delta),f'(x)>0 \\ \end{matrix} \right\}\Rightarrow f(x)\text{在}x=x_0\text{处取得极小值}

若左邻域,右邻域,一阶导数符号不变,f(x)在x=x0处不取极值\text{若左邻域,右邻域,一阶导数符号不变,}f(x)\text{在}x=x_0\text{处不取极值}

定理三

设f(x)在x=x0处f′(x)=0,f′′(x0)存在f′′(x0)≠0\text{设}f(x)\text{在}x=x_0\text{处}f'(x)=0,f''(x_0)\text{存在}f''(x_0)\ne0

f′′(x0)<0⇔f(x)在x0处取得极大值f''(x_0)<0\Leftrightarrow f(x)\text{在}x_0\text{处取得极大值}

f′′(x0)>0⇔f(x)在x0处取得极小值f''(x_0)>0\Leftrightarrow f(x)\text{在}x_0\text{处取得极小值}

最值

f(x)在U上有定义,∃x0∈U,使得∀x∈U都有:f(x)\text{在}U\text{上有定义},\exists x_0\in U,\text{使得}\forall x\in U\text{都有}\text{:}

f(x)≤f(x0)⇒f(x0)为f(x)在U上的最大值f(x)\leq f(x_0)\Rightarrow f(x_0)\text{为}f(x)\text{在}U\text{上的最大值}

f(x)≥f(x0)⇒f(x0)为f(x)在U上的最大值f(x)\geq f(x_0)\Rightarrow f(x_0)\text{为}f(x)\text{在U上的最大值}

最值不一定是极值,极值不一定是最值

不等式链

21x+1y≤xy≤x+y2≤x2+y22\frac{2}{\frac{1}{x}+\frac{1}{y}}\leq\sqrt{xy}\leq\frac{x+y}{2}\leq\sqrt{\frac{x^2+y^2}{2}}

例题

例1

m,n>0且m+n−23=0,则max⁡(mn)为?m,n>0\text{且}m+n-2\sqrt{3}=0,\text{则}\max(mn)\text{为?}

解法一:[基本不等式]\text{解法一}:[\text{基本不等式}]

m+n=23m+n=2\sqrt{3}

mn≤m+n2mn≤(m+n2)2mn≤(232)2mn≤3当且仅当m=n时取等max(mn)=3\begin{equation} \begin{aligned} \sqrt{mn}&\leq\frac{m+n}{2}\\ mn&\leq(\frac{m+n}{2})^2\\ mn&\leq(\frac{2\sqrt{3}}{2})^2\\ mn&\leq3\\ \text{当且仅当}m=n\text{时取等}\\ max(mn)=3 \end{aligned} \end{equation}


解法二:[拉格朗日数乘]\text{解法二}:[\text{拉格朗日数乘}]

使用前提:函数最值存在,若函数本身不存在最值,则此法的解皆不是最值约束条件没有边界约束函数需具有一阶连续偏导数,且约束函数的梯度向量不为0\begin{aligned} & \text{使用前提}: \\ & \text{函数最值存在,若函数本身不存在最值,则此法的解皆不是最值}\\ & \text{约束条件没有边界}\\ & \text{约束函数需具有一阶连续偏导数,且约束函数的梯度向量不为}0 \end{aligned}

列式联立方程解得f=mng(x)=0λ=3g=m+n−23=0∂h∂m=0m=3h=f−λg∂h∂n=0n=3\begin{array}{c:c:c} \text{列式} & \text{联立方程}&\text{解得}\\ f=mn&g(x)=0&\lambda = \sqrt{3}\\ g=m+n-2\sqrt{3}=0&\frac{\partial h}{\partial m}=0&m= \sqrt{3}\\ h=f-\lambda g&\frac{\partial h}{\partial n}=0&n = \sqrt{3}\\ \end{array}