多元正常后验
这是一个非常简单的问题,但我无法在互联网上或书中的任何地方找到推导。我想看到一个贝叶斯如何更新多元正态分布的推导。例如:想象一下 P(x|μ,Σ)P(μ)==N(μ,Σ)N(μ0,Σ0).P(x|μ,Σ)=N(μ,Σ)P(μ)=N(μ0,Σ0). \begin{array}{rcl} \mathbb{P}({\bf x}|{\bf μ},{\bf Σ}) & = & N({\bf \mu}, {\bf \Sigma}) \\ \mathbb{P}({\bf \mu}) &= & N({\bf \mu_0}, {\bf \Sigma_0})\,. \end{array} 观察一组x1...xnx1...xn{\bf x_1 ... x_n},我想计算P(μ|x1...xn)P(μ|x1...xn)\mathbb{P}({\bf \mu | x_1 ... x_n})。我知道答案是P(μ|x1...xn)=N(μn,Σn)P(μ|x1...xn)=N(μn,Σn)\mathbb{P}({\bf \mu | x_1 ... x_n}) = N({\bf \mu_n}, {\bf \Sigma_n})其中 μnΣn==Σ0(Σ0+1nΣ)−1(1n∑i=1nxi)+1nΣ(Σ0+1nΣ)−1μ0Σ0(Σ0+1nΣ)−11nΣμn=Σ0(Σ0+1nΣ)−1(1n∑i=1nxi)+1nΣ(Σ0+1nΣ)−1μ0Σn=Σ0(Σ0+1nΣ)−11nΣ \begin{array}{rcl} \bf \mu_n &=& \displaystyle\Sigma_0 \left(\Sigma_0 …