Efficient sampling from a multivariate Gaussian Mixture distribution for a given CDF level

52 Views Asked by At

I have a multivariate Gaussian Mixture (GM) distribution. I am wondering if there is any more efficient way of drawing samples (i.e., identify the iso-surface) from a multivariate Gaussian Mixture distribution for a given joint CDF (joint cumulative distribution function) level?

Here is my current solution (a brutal one). I'm generating a large number of realizations from the GM distribution and evaluate the corresponding CDF value of each of the realizations. Then I select those that are closest to the target CDF value. But this method can be quite time consuming because the large number of CDF evaluations.

Thanks for your help!