INLA feature request

[arose13]

I don't know if this is in PyMC3 scope but I was thinking since we have variational inference perhaps we could consider adding INLA

You can find a brief intro is here

[springcoil]

Has anyone any ideas how hard this would be to implement?

[junpenglao]

I did some prototype here: https://github.com/junpenglao/Planet_Sakaar_Data_Science/blob/master/Ports/Laplace%20approximation%20in%20pymc3.ipynb

[arose13]

@junpenglao your notebook is fantastic. It's helping me feel the background material.

I also should I/we follow the .aprrox() API? In R the API is a formula (patsy like API) which I feel may not be appropriate.

[springcoil]

The prototype is fairly involved :)

[ricardoV94]

Closing due to lack of activity. If anyone wants to take on this future, it should probably go into https://github.com/pymc-devs/pymc-experimental