Ch02

ZhuYuanxiang 2023-10-20 10:22:24
Categories: Tags:

C02. Bayes 决策论

2.1 引言

Bayes 决策论是对折衷的量化,这个折衷源于使用概率进行不同分类决策与这种决策伴随的相应的代价。

Bayes 公式

判别条件

2.2 Bayes 决策论——连续特征

前提条件

推导过程

实际案例:二分类问题

$$
\begin{aligned}
R(\omega_1|\text{x})&=\lambda_{11} P(\omega_1|\text{x})+\lambda_{12} P(\omega_2|\text{x})\
R(\omega_2|\text{x})&=\lambda_{21} P(\omega_1|\text{x})+\lambda_{22} P(\omega_2|\text{x})
\end{aligned}
$$

$$
\begin{aligned}
(\lambda_{21}-\lambda_{11})P(\omega_1|\text{x})&>(\lambda_{12}-\lambda_{22})P(\omega_2|\text{x})\
(\lambda_{21}-\lambda_{11})p(\text{x}|\omega_1)P(\omega_1)&>(\lambda_{12}-\lambda_{22})p(\text{x}|\omega_2) P(\omega_2)\
\frac{p(\text{x}|\omega_1)}{p(\text{x}|\omega_2)}&>\frac{\lambda_{12}-\lambda_{22}}{\lambda_{21}-\lambda_{11}} \frac{P(\omega_2)}{P(\omega_1)}
\end{aligned}
$$

2.3 最小误差率分类