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
The MDN i.e. Gaussian Mixture Density network, is a density estimator which gets a lot of "special" treatment.
SNPE_A requires it, and implements a special "corrected" SNPE_A_MDN wrapper.
SNPE_C has a special loss for it
conditional_density.py implements a special ConditionedMDN to use the "marginalization" properties.
Hence it may be good to collapse all current MDN wrappers/functionalities into a single class MDNDensityEstimator, which does follow the new DensityEstimator.
Advantages:
Reducing code duplication.
Better typing.
The text was updated successfully, but these errors were encountered:
I find the SNPE_A_MDN class quite confusing. Is there a loss implemented for it yet? Why does it only inherit from DensityEstimator but not implement all methods?
Description
The MDN i.e. Gaussian Mixture Density network, is a density estimator which gets a lot of "special" treatment.
SNPE_A_MDN
wrapper.conditional_density.py
implements a specialConditionedMDN
to use the "marginalization" properties.Hence it may be good to collapse all current MDN wrappers/functionalities into a single class
MDNDensityEstimator
, which does follow the newDensityEstimator
.Advantages:
The text was updated successfully, but these errors were encountered: