基本概念
本片博文为后续博文做基础铺垫,给出符号和基本公式的定义:
- 随机变量
X
X
X的值为
x
x
x:
p
(
X
=
x
)
p(X=x)
p(X=x) 简写为
p
(
x
)
p(x)
p(x)
∑ x p ( x ) = 1 离散 ∫ p ( x ) d x = 1 连续 (1.1) \sum_xp(x) = 1 ~~~~~~~~~\text{离散}\\ \int p(x)dx=1 ~~~~~~~\text{连续}\tag{1.1} x∑p(x)=1 离散∫p(x)dx=1 连续(1.1) - 正态分布
N
(
x
;
u
,
σ
2
)
N(x;u,\sigma^2)
N(x;u,σ2):
p ( x ) = ( 2 π σ 2 ) − 1 2 exp { − 1 2 ( x − μ ) 2 σ 2 } ↓ 多维 p ( x ) = d e t ( 2 π Σ 2 ) − 1 2 exp { − 1 2 ( x − μ ) T Σ − 1 ( x − μ ) } (1.2) p(x) = (2\pi\sigma^2)^{-\frac{1}{2}}\exp\left \{ -\frac{1}{2} \frac{(x-\mu)^2}{\sigma^2}\right \}\\ \downarrow \text{多维}\\ p(x) = det(2\pi\Sigma^2)^{-\frac{1}{2}}\exp\left \{ -\frac{1}{2} (x-\mu)^T\Sigma^{-1}(x-\mu)\right \}\tag{1.2} p(x)=(2πσ2)−21exp{−21σ2(x−μ)2}↓多维p(x)=det(2πΣ2)−21exp{−21(x−μ)TΣ−1(x−μ)}(1.2) - 联合分布:
p ( x , y ) = p ( X = x ∪ Y = y ) ↓ X , Y 独立 p ( x , y ) = p ( x ) p ( y ) (1.3) p(x,y) = p(X=x \cup Y=y)\\ \downarrow X,Y\text{独立}\\ p(x,y) = p(x)p(y)\tag{1.3} p(x,y)=p(X=x∪Y=y)↓X,Y独立p(x,y)=p(x)p(y)(1.3) - 条件分布:
p ( x ∣ y ) = p ( X = x ∣ Y = y ) ↓ p ( x ∣ y ) = p ( x , y ) p ( y ) ↓ X , Y 独立 p ( x ∣ y ) = p ( x ) p ( y ) p ( y ) = p ( x ) (1.4) p(x|y) = p(X=x|Y=y)\\ \downarrow\\ p(x|y) =\frac{p(x,y)}{p(y)}\\ \downarrow X,Y\text{独立}\\ p(x|y) =\frac{p(x)p(y)}{p(y)} = p(x)\tag{1.4} p(x∣y)=p(X=x∣Y=y)↓p(x∣y)=p(y)p(x,y)↓X,Y独立p(x∣y)=p(y)p(x)p(y)=p(x)(1.4) - 全概率公式:
p ( x ) = ∑ y p ( x ∣ y ) p ( y ) 离散 p ( x ) = ∫ p ( x ∣ y ) p ( y ) d y 连续 (1.5) p(x) = \sum_y p(x|y)p(y) ~~~~~~ \text{离散}\\ p(x) = \int p(x|y)p(y)dy ~~~~~~ \text{连续}\tag{1.5} p(x)=y∑p(x∣y)p(y) 离散p(x)=∫p(x∣y)p(y)dy 连续(1.5) - 贝叶斯准则:(很重要)
p ( x ∣ y ) = p ( y ∣ x ) p ( x ) p ( y ) = p ( y ∣ x ) p ( x ) ∑ x ′ p ( y ∣ x ′ ) p ( x ′ ) 离散 p ( x ∣ y ) = p ( y ∣ x ) p ( x ) p ( y ) = p ( y ∣ x ) p ( x ) ∫ p ( y ∣ x ′ ) p ( x ′ ) d x ′ 连续 ↓ p ( y ) − 1 被写为 η p ( x ∣ y ) = η p ( y ∣ x ) p ( x ) (1.6) p(x|y) =\frac{p(y|x)p(x)}{p(y)} = \frac{p(y|x)p(x)}{\sum_{x'}p(y|x')p(x')} ~~~~~~ \text{离散}\\ p(x|y) =\frac{p(y|x)p(x)}{p(y)} = \frac{p(y|x)p(x)}{\int p(y|x')p(x')dx'} ~~~~~~ \text{连续}\\ \downarrow p(y)^{-1}\text{被写为}\eta\\ p(x|y) = \eta p(y|x)p(x)\tag{1.6} p(x∣y)=p(y)p(y∣x)p(x)=∑x′p(y∣x′)p(x′)p(y∣x)p(x) 离散p(x∣y)=p(y)p(y∣x)p(x)=∫p(y∣x′)p(x′)dx′p(y∣x)p(x) 连续↓p(y)−1被写为ηp(x∣y)=ηp(y∣x)p(x)(1.6) - 多变量条件概率:
p ( x ∣ y , z ) = p ( y ∣ x , z ) p ( x ∣ z ) p ( y ∣ z ) (1.7) p(x|y,z) =\frac{p(y|x,z)p(x|z)}{p(y|z)}\tag{1.7} p(x∣y,z)=p(y∣z)p(y∣x,z)p(x∣z)(1.7) - 条件概率的独立性(X,Y独立):
p ( x , y ∣ z ) = p ( x ∣ z ) p ( y ∣ z ) p ( x ∣ z ) = p ( x ∣ z , y ) p ( y ∣ z ) = p ( y ∣ z , x ) (1.8) p(x,y|z) = p(x|z)p(y|z)\\ p(x|z) = p(x|z,y)\\ p(y|z) = p(y|z,x)\tag{1.8} p(x,y∣z)=p(x∣z)p(y∣z)p(x∣z)=p(x∣z,y)p(y∣z)=p(y∣z,x)(1.8) - 条件独立不一定绝对独立,绝对独立不一定条件独立:
p ( x , y ∣ z ) = p ( x ∣ z ) p ( y ∣ z ) ⇏ p ( x , y ) = p ( x ) p ( y ) p ( x , y ) = p ( x ) p ( y ) ⇏ p ( x , y ∣ z ) = p ( x ∣ z ) p ( y ∣ z ) (1.9) \begin{aligned} p(x,y|z) = p(x|z)p(y|z) &\nRightarrow p(x,y) = p(x)p(y)\\ p(x,y) = p(x)p(y) &\nRightarrow p(x,y|z) = p(x|z)p(y|z)\tag{1.9} \end{aligned} p(x,y∣z)=p(x∣z)p(y∣z)p(x,y)=p(x)p(y)⇏p(x,y)=p(x)p(y)⇏p(x,y∣z)=p(x∣z)p(y∣z)(1.9) - 期望:
E [ X ] = ∑ x x p ( x ) 离散 E [ X ] = ∫ x p ( x ) d x 连续 ↓ E [ a X + b ] = a E [ X ] + b (1.10) E[X] = \sum_x xp(x) ~~~~~~ \text{离散}\\ E[X] = \int xp(x)dx ~~~~~~ \text{连续}\\\tag{1.10} \downarrow\\ E[aX + b] = aE[X] +b E[X]=x∑xp(x) 离散E[X]=∫xp(x)dx 连续↓E[aX+b]=aE[X]+b(1.10) - 协方差:
C o v [ X ] = E [ X − E [ X ] ] 2 = E [ X 2 ] − E [ X ] 2 (1.11) Cov[X] = E[X-E[X]]^2 = E[X^2] - E[X]^2\tag{1.11} Cov[X]=E[X−E[X]]2=E[X2]−E[X]2(1.11) - 熵:
H p ( x ) = E [ − log 2 p ( x ) ] ↓ H p ( x ) = − ∑ x p ( x ) log 2 p ( x ) 离散 H p ( x ) = ∫ p ( x ) log 2 p ( x ) d x 连续 (1.12) H_p(x) = E[-\log_2p(x)]\\ \downarrow\\ H_p(x) = -\sum_x p(x)\log_2p(x) ~~~~~~ \text{离散}\\ H_p(x) = \int p(x)\log_2p(x)dx ~~~~~~ \text{连续}\\\tag{1.12} Hp(x)=E[−log2p(x)]↓Hp(x)=−x∑p(x)log2p(x) 离散Hp(x)=∫p(x)log2p(x)dx 连续(1.12) - 测量值
z
z
z:
z t 1 : t 2 = z t 1 , z t 1 + 1 , z t 1 + 2 , ⋯ , z t 2 (1.13) z_{t_1:t_2} = z_{t_1}, z_{t_1+1}, z_{t_1+2}, \cdots, z_{t_2}\tag{1.13} zt1:t2=zt1,zt1+1,zt1+2,⋯,zt2(1.13) - 控制值
u
u
u:
u t 1 : t 2 = u t 1 , u t 1 + 1 , u t 1 + 2 , ⋯ , u t 2 (1.14) u_{t_1:t_2} = u_{t_1}, u_{t_1+1}, u_{t_1+2}, \cdots, u_{t_2}\tag{1.14} ut1:t2=ut1,ut1+1,ut1+2,⋯,ut2(1.14) - 条件独立性的例子:
- 状态变量 x t x_t xt仅与状态 x t − 1 x_{t-1} xt−1和控制量 u t u_t ut有关
- 观测值
z
t
z_t
zt仅与状态
x
t
x_{t}
xt有关
p ( x t ∣ x 0 : t − 1 , z 1 : t − 1 , u 1 : t ) = p ( x t ∣ x t − 1 , u t ) p ( z t ∣ x 0 : t , z 1 : t − 1 , u 1 : t ) = p ( z t ∣ x t ) (1.15) p(x_t|x_{0:t-1},z_{1:t-1},u_{1:t}) = p(x_t|x_{t-1},u_t) p(z_t|x_{0:t},z_{1:t-1},u_{1:t}) = p(z_t|x_{t})\tag{1.15} p(xt∣x0:t−1,z1:t−1,u1:t)=p(xt∣xt−1,ut)p(zt∣x0:t,z1:t−1,u1:t)=p(zt∣xt)(1.15)
- 置信度:
b e l ( x t ) = p ( x t ∣ z 1 : t , u 1 : t ) b e l ˉ ( x t ) = p ( x t ∣ z 1 : t − 1 , u 1 : t ) (1.16) bel(x_t) = p(x_t|z_{1:t},u_{1:t})\\ \bar{bel}(x_t) = p(x_t|z_{1:t-1},u_{1:t})\tag{1.16} bel(xt)=p(xt∣z1:t,u1:t)belˉ(xt)=p(xt∣z1:t−1,u1:t)(1.16)
从 b e l ˉ ( x t ) \bar{bel}(x_t) belˉ(xt)中计算 b e l ( x t ) bel(x_t) bel(xt)称为修正或测量更新
下一章
- 下一章讲:贝叶斯滤波算法
参考文章
- Thrun, & Sebastian. (2005). Probabilistic robotics.