机器学习7 拉格朗日乘数法

最新推荐文章于 2022-10-07 15:59:54 发布

skmygdrs

最新推荐文章于 2022-10-07 15:59:54 发布

阅读量391

点赞数 1

分类专栏： machine_learning

本文链接：https://blog.csdn.net/skmygdrs/article/details/52101183

版权

machine_learning 专栏收录该内容

9 篇文章 0 订阅

订阅专栏

{z = f (x, y) h (x, y) = 0} \to min f (x, y) h (x, y) = 0 \Rightarrow y = φ (x) \partial h ( x , y ) \partial x + \partial h ( x , y ) \partial y d ( φ ( x ) ) d ( x ) = 0 f (x, y) = f (x, φ (x)) \Rightarrow \partial f ( x , y ) \partial x + \partial f ( x , y ) \partial y d ( φ ( x ) ) d ( x ) = 0 \Leftrightarrow ⎧ ⎩ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ h (x, y) = 0 \partial h ( x , y ) \partial x + \partial h ( x , y ) \partial y d ( φ ( x ) ) d ( x ) = 0 \partial f ( x , y ) \partial x + \partial f ( x , y ) \partial y d ( φ ( x ) ) d ( x ) = 0 \Leftrightarrow ⎧ ⎩ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ h (x, y) = 0 \partial f ( x , y ) \partial x \partial h ( x , y ) \partial x = \partial f ( x , y ) \partial y \partial h ( x , y ) \partial y = - λ \Leftrightarrow ⎧ ⎩ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ h (x, y) = 0 \partial f ( x , y ) \partial x + λ \partial h ( x , y ) \partial x = 0 \partial f ( x , y ) \partial y + λ \partial h ( x , y ) \partial y = 0 L (x, y, λ) = f (x, y) + λ h (x, y) \Leftrightarrow ⎧ ⎩ ⎨ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ \partial L ( x , y ) \partial λ = 0 \partial L ( x , y ) \partial x = 0 \partial L ( x , y ) \partial y = 0

$\eqalign{ & \left\{ \matrix{ z = f\left( {x,y} \right) \cr h\left( {x,y} \right) = 0 \cr} \right\} \to \min f(x,y) \cr & h\left( {x,y} \right) = 0 \Rightarrow y = \varphi \left( x \right) \cr & {{\partial h\left( {x,y} \right)} \over {\partial x}} + {{\partial h\left( {x,y} \right)} \over {\partial y}}{{d\left( {\varphi \left( x \right)} \right)} \over {d\left( x \right)}} = 0 \cr & f\left( {x,y} \right) = f\left( {x,\varphi \left( x \right)} \right){\kern 1pt} \cr & \Rightarrow \cr & {{\partial f\left( {x,y} \right)} \over {\partial x}} + {{\partial f\left( {x,y} \right)} \over {\partial y}}{{d\left( {\varphi \left( x \right)} \right)} \over {d\left( x \right)}} = 0 \cr & \Leftrightarrow \cr & \left\{ \matrix{ h\left( {x,y} \right) = 0 \cr {{\partial h\left( {x,y} \right)} \over {\partial x}} + {{\partial h\left( {x,y} \right)} \over {\partial y}}{{d\left( {\varphi \left( x \right)} \right)} \over {d\left( x \right)}} = 0 \cr {{\partial f\left( {x,y} \right)} \over {\partial x}} + {{\partial f\left( {x,y} \right)} \over {\partial y}}{{d\left( {\varphi \left( x \right)} \right)} \over {d\left( x \right)}} = 0 \cr} \right. \cr & \Leftrightarrow \cr & \left\{ \matrix{ h\left( {x,y} \right) = 0 \cr {{{{\partial f\left( {x,y} \right)} \over {\partial x}}} \over {{{\partial h\left( {x,y} \right)} \over {\partial x}}}}{\rm{ = }}{{{{\partial f\left( {x,y} \right)} \over {\partial y}}} \over {{{\partial h\left( {x,y} \right)} \over {\partial y}}}} = - \lambda \cr} \right. \cr & \Leftrightarrow \cr & \left\{ \matrix{ h\left( {x,y} \right) = 0 \cr {{\partial f\left( {x,y} \right)} \over {\partial x}} + \lambda {{\partial h\left( {x,y} \right)} \over {\partial x}} = 0 \cr {{\partial f\left( {x,y} \right)} \over {\partial y}} + \lambda {{\partial h\left( {x,y} \right)} \over {\partial y}} = 0 \cr} \right. \cr & L(x,y,\lambda ) = f\left( {x,y} \right) + \lambda h\left( {x,y} \right) \cr & \Leftrightarrow \cr & \left\{ \matrix{ {{\partial L\left( {x,y} \right)} \over {\partial \lambda }} = 0 \cr {{\partial L\left( {x,y} \right)} \over {\partial x}} = 0 \cr {{\partial L\left( {x,y} \right)} \over {\partial y}} = 0 \cr} \right. \cr & \cr}$

skmygdrs

关注

1
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
机器学习7 拉格朗日乘数法

{z=f(x,y)h(x,y)=0}→minf(x,y)h(x,y)=0⇒y=φ(x)∂h(x,y)∂x+∂h(x,y)∂yd(φ(x))d(x)=0f(x,y)=f(x,φ(x))⇒∂f(x,y)∂x+∂f(x,y)∂yd(φ(x))d(x)=0⇔⎧⎩⎨⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪h(x,y)=0∂h(x,y)∂x+∂h(x,y)∂yd(φ(x))d(x)=0∂f(x,y)∂x+∂f(x,y)∂yd(
复制链接

扫一扫