function M = sample_gaussian(mu, Sigma, N)
mu = mu(:);
n=length(mu);
[U,D,V] = svd(Sigma);
M = randn(n,N);
M = (U*sqrt(D))*M + mu*ones(1,N);
M = M';
Subscribe to:
Post Comments (Atom)
About Me
- Yi Wang
- Beijing, China
- I am an engineer working on large-scale parallel machine learning and data mining. This Blog records my learning& working notes. For a Blog on my daily life, you may want to have a look at http://cxwangyi.spaces.msn.com.
3 comments:
Dear Yi Wang,
we were experiencing problems with Matlab slicesample function when sampling from a 3-variate Gaussian distribution with mean=[0 0 0] and covariance matrix something like this cov=diag(1e-7, 1e-7, 10). Slicesample seems to fail when samples the third dimension scattering the samples with a very tight variance (not 10, but more like 0.6).
Due to this limitation of slicesample we searched in the web and found your code. And it works great!!! Thanks!
Do you have any paper or notes explaining the method you used?
Thanks
Andrea
Yes,
great piece of code.
Could you tell us what is the theory behind it?
best,
Joe
Hi, Andrea and Joe,
I am sorry for the late response. I did not get Gmail notification about your comments .... do not know why.
I got the idea from this paper: Sampling from the Multivariate Gaussian Distribution using Reconfigurable Hardware, which refers to an even older paper: A comparison of multivariate normal generators.
Post a Comment