Implementation and derivation of "Variational Bayesian inference for a nonlinear forward model." [Chappell et al. 2008] for arbitrary, user-defined model errors.
Those two distributions are tailored for our VB algorithm but the associated values -- means and precision matrix from vb.MVN and shape and scale from vb.Gamma -- can also be extracted from scipy.stats.multivariate_normal and scipy.stats.gamma.
IMO the only question would be how to transfer the feature of vb.MVN.named_mean("some_prm_name") there...
Over time, the distributions gained a lot of functionality related to post-processing and creation. Their methods dist allow for conversion to the scipy distributions.
Those two distributions are tailored for our VB algorithm but the associated values -- means and precision matrix from vb.MVN and shape and scale from vb.Gamma -- can also be extracted from scipy.stats.multivariate_normal and scipy.stats.gamma.
IMO the only question would be how to transfer the feature of
vb.MVN.named_mean("some_prm_name")
there...