Gaussianmixture
This allows notebook to run on scikit-learn 0.18+. The calling convention has changed since that release, with name GaussianMixture instead of old GMM. Parameter covariances_ instead of covars_, random_state can only be assigned to model, not the sample, and sample now returns a tuple of which the 0th element is the desired array of samples. Also fixed a deprecation warning about make_blobs which is now to be imported directly from sklearn.datasets instead of deprecated samples_generator. I've tested this on my Anaconda install (Python-3.8.5, scikit-learn-0.23.2) and on colab. (Thanks for producing this fantastic book and this demo!)
Hi @jakevdp,
I think this PR will be helpful if merged. I recently encountered the problems resolved by this PR.