0

As in the ScikitLearn GaussianMixture model, reg_covar=1e-06 adds Non-negative regularization to the diagonal of covariance, which ensures covariance matrices are positive definite.

sklearn.mixture.GaussianMixture(n_components=1, *, covariance_type='full', tol=0.001, reg_covar=1e-06, max_iter=100, n_init=1, init_params='kmeans', weights_init=None, means_init=None, precisions_init=None, random_state=None, warm_start=False, verbose=0, verbose_interval=10)[source]

How to set this kind of constraint to get positive definite covariance matrices in pomegranate MultivariateGaussianDistribution.


from sklearn import datasets
iris = datasets.load_iris()
import pomegranate
pomegranate.gmm.GeneralMixtureModel.from_samples(pomegranate.MultivariateGaussianDistribution, n_components=3, X=iris.data)

Thank you for your valuable time.

iforcebd
  • 87
  • 11
  • Do you have any ideas @StupidWolf – iforcebd Apr 17 '21 at 13:11
  • ok, I added dataset importing code but my question is how can we make sure covariance matrices are positive definite e.g. reg_covar=1e-06 in scikit-learn Gaussian mixture model? because when I apply it in my project up to 30 topics its works fine, but for 60 topics it giving error message: covariance matrices are not positive definite(asymmetric). @ D Adams – iforcebd Apr 19 '21 at 06:25

0 Answers0