python - Sampling parts of a vector from gaussian mixture model -

python - Sampling parts of a vector from gaussian mixture model -

i want sample elements of vector sum of gaussians given means , covariance matrices.

specifically:

i'm imputing info using gaussian mixture model (gmm). i'm using next procedure , sklearn:

impute mean get means , covariances gmm (for illustration 5 components) take 1 of samples , sample missing values. other values remain same. repeat few times

there 2 problems see this. (a) how sample sum of gaussians, (b) how sample part of vector. assume both can solved @ same time. (a), can utilize rejection sampling or inverse transform sampling sense there improve way utilizing multivariate normal distribution generators in numpy. or, other efficient method. (b), need multiply sampled variable gaussian has known values sample argument. right?

i prefer solution in python algorithm or pseudocode sufficient.

since sampling relative proportion of distribution matters, scaling preface or can thrown away. diagonal covariance matrix, 1 can utilize covariance submarine , mean subvector has dimensions of missing data. covariance off-diagonal elements, mean , std dev of sampling gaussian need changed.

python numpy random-sample normal-distribution mixture-model