Ismail Senoz

Problems with Laplace approximation method

3

There are couple of problems with the Laplace approximation: 1. For Laplace approximation we use Optim.jl package to evaluate derivatives, however in current version of Optim.jl derivative syntax for 1D...

Sometimes we send pdf messages and need to evaluate the expectation statistics of these un-normalized pdf messages. Under the assumption that these messages are integrable, we resort to importance sampling...

Currently autoupdates do not support updating when statistics are indexed by `end` as in the example ``` update = @autoupdates begin mx = mean(q(x))[1:end-1] Vx = cov(q(x))[1:end-1, 1:end-1] end ```...