HMC models

[dPys]

@arokem -- are there any other models in dipy besides tensor, sfm, csd, and maybe csa that could be worth including?

As for SFM, what about using the BayesianRidge solver? ... i.e. where iterations of the entire LOO procedure gradually build a posterior on alpha/lambda?

Alternatively, or in addition, we might try building a generic subject-movement model using a group of subjects and use its predictions as an initializer for the subject-specific model?

[arokem]

are there any other models in dipy besides tensor, sfm, csd, and maybe csa that could be worth including?

I would not include CSD/CSA. These are models primarily focused on describing the ODF, and not on providing an accurate signal reconstrution. I would definitely include DKI.

As for SFM, what about using the BayesianRidge solver? ... i.e. where iterations of the entire LOO procedure gradually build a posterior on alpha/lambda?

I like that idea!

[oesteban]

I concur with Ariel's suggestions.

Let's open up new issues for each model we want to implement, and another one for the BayesianRidge solver because it feels like a very promising suggestion.