You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Building on Marcel's work, add an MDN (Mixture Density Network) implementation that can play well with e.g. SNPE-A proposal posterior corrections.
This is the wishlist
basic MDN that works and is useable from sbi (mixture of full-covariance Gaussians)
MDN becomes highly compositional - only last layer implemented (MDN layer), the rest is built e.g. using Sequential and possibly heuristics given data dimensions, etc.
MDN layer returns a MixtureSameFamily density (becoming part of PyTorch, PR)
(long term) MDN layer can self-configure given the particulars of the desired MixtureSameFamily
Building the whole MDN does not require specification of redundant information, much like Keras s sequential (or look at thinc for a more functional take).
The text was updated successfully, but these errors were encountered:
Building on Marcel's work, add an MDN (Mixture Density Network) implementation that can play well with e.g. SNPE-A proposal posterior corrections.
This is the wishlist
Sequential
and possibly heuristics given data dimensions, etc.MixtureSameFamily
density (becoming part of PyTorch, PR)MixtureSameFamily
The text was updated successfully, but these errors were encountered: