机器学习5 多维正态分布(高斯分布)

概率密度

$p(x) = {1 \over {{{\left( {2\pi } \right)}^{{n \over 2}}}{{\left| \beta \right|}^{{1 \over 2}}}}}\exp \left( { - {1 \over 2}{{\left( {x - \alpha } \right)}^T}{\beta ^{ - 1}}(x - \alpha )} \right)$~$N\left( {\alpha ,\beta } \right)$
其中 $\beta = E\left[ {\left( {x - \alpha } \right){{\left( {x - \alpha } \right)}^T}} \right]$
当n=1时
令 $\beta = {\sigma ^2},\alpha = \mu$
则 $p(x) = {1 \over {\sqrt {2\pi } \sigma }}\exp \left( { - {{{{\left( {x - \mu } \right)}^2}} \over {2{\sigma ^2}}}} \right)$
当n=1时
令

$$\beta = \left( {\matrix{ {{\sigma _1}^2} & {{\sigma _1}{\sigma _2}r} \cr {{\sigma _1}{\sigma _2}r} & {{\sigma _2}^2} \cr } } \right)$$
$$\alpha = \left( {a,b} \right)$$
则
${\beta ^{ - 1}} = {1 \over {{\sigma _1}^2{\sigma _2}^2\left( {1 - {r^2}} \right)}}\left( {\matrix{ {{\sigma _2}^2} & { - {\sigma _1}{\sigma _2}r} \cr { - {\sigma _1}{\sigma _2}r} & {{\sigma _1}^2} \cr } } \right)$
$p(x,y) = {1 \over {2\pi {\sigma _1}{\sigma _2}\sqrt {1 - {r^2}} }}\exp \left( { - {1 \over {2\left( {1 - {r^2}} \right)}}\left[ {{{{{\left( {x - a} \right)}^2}} \over {{\sigma _1}^2}} - {{2r\left( {x - a} \right)\left( {y - b} \right)} \over {{\sigma _1}{\sigma _2}}} + {{{{\left( {y - b} \right)}^2}} \over {{\sigma _2}^2}}} \right]} \right)$

二维边缘分布率

$\int_{ - \infty }^\infty {p(x,y)} dy =$
$\int_{ - \infty }^\infty {{1 \over {2\pi {\sigma _1}{\sigma _2}\sqrt {1 - {r^2}} }}\exp \left( { - {1 \over {2\left( {1 - {r^2}} \right)}}\left[ {{{{{\left( {x - a} \right)}^2}} \over {{\sigma _1}^2}} - {{2r\left( {x - a} \right)\left( {y - b} \right)} \over {{\sigma _1}{\sigma _2}}} + {{{{\left( {y - b} \right)}^2}} \over {{\sigma _2}^2}}} \right]} \right)} dy$
$= \int_{ - \infty }^\infty {{1 \over {\sqrt {2\pi } {\sigma _1}}}{1 \over {\sqrt {2\pi } {\sigma _2}\sqrt {1 - {r^2}} }}\exp \left( { - {1 \over {2\left( {1 - {r^2}} \right)}}\left[ {{{{{\left( {y - b - {{{\sigma _2}r\left( {x - a} \right)} \over {{\sigma _1}}}} \right)}^2}} \over {{\sigma _2}^2}} + \left( {1 - {r^2}} \right){{{{\left( {x - a} \right)}^2}} \over {{\sigma _1}^2}}} \right]} \right)} dy$
$= {1 \over {\sqrt {2\pi } {\sigma _1}}}\exp \left( { - {{{{\left( {x - a} \right)}^2}} \over {2{\sigma _1}^2}}} \right)$