Open rainwoodman opened 7 years ago
The EM algorithm for Gaussian mixtures is so simple, we shouldn't need to import anything. Especially in 2D, where the matrix computations are essentially trivial. For example look at: http://www.ics.uci.edu/~smyth/courses/cs274/notes/EMnotes.pdf
I suppose projection and EM do not commute. Not a concern for visualization?
And compare it with sklearn.mixture.GMM.
The idea is we shall be able to run without importing sklearn.