I have two GMM distribution (K=2) of multi-variant Gaussian distributions whose parameters(mean and sigma vectors and weight)are found using the EM algorithm. How can I find the the joint distribution of the two GMMs?
Joint Distribution of Two Gaussian Mixture Model
0
$\begingroup$
probability
probability-theory
normal-distribution
machine-learning
-
0I think the joint distribution of the two GMMs is just a GMM distribution with $K=4$. You can assume that there is a prior probability of each GMM. – 2017-02-07
-
0thanks Jy. now,how to i do that? have you a mathematical equation? – 2017-02-08
1 Answers
0
Suppose that the density of 2 GMMs are $f_1(x)=\pi_1\phi_{\theta_{1,1}}(x) + (1-\pi_1) \phi_{\theta_{1,2}}(x)$ and $f_2(x)=\pi_2\phi_{\theta_{2,1}}(x) + (1-\pi_2) \phi_{\theta_{2,2}}(x)$ respectively.
Let $p$ be the prior probability of $x$ generated by GMM1. Then the joint density is $$f(x)=p*f_1(x)+(1-p)*f_2(x)$$ which is the density of a GMM with $K = 4$ since $$p\pi_1 + p(1-\pi_1) + (1-p)\pi_2 + (1-p)(1-\pi_2)=1$$