Last active
October 18, 2019 11:11
-
-
Save junpenglao/f5b3749393c92f516230a1ea389d1380 to your computer and use it in GitHub Desktop.
Documents/Github/Human_Learning/Miscellaneous/[WIP] Bayesian GMM.ipynb
It should now works a bit more naturally after pymc-devs/pymc#2904 - there is no need to write down the custom logp any more.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Currently, this is not working too well yet. To improve from here, it might be necessary to follow Kamary, Lee & Robert (2017), or the more classical Expectation Propagation. It doesn't hurt also to revisited the relevant chapters in Christopher Bishop's Pattern Recognition and Machine Learning.
Better sampling/fitting could be archived by utilising the composing inferences in PyMC3 or Edward (by implementing the special models as described in the papers mentioned above).