Hi Guys -

Trying out DPGMM to see if I can get a more stable solution. Things look good so far except for one problem.

The sample() method does not work for dpgmm. The primary reason is that sample() assumes there is a self.covars_, but this does not exist.

There is a secondary problem however. I tried using get_covars to set the covars but these covariances are not correct, perhaps because of a different convention somehow in the definitions of these covariances.

best, and thanks for the good work on sklearn,
-e