Need for MvNormal with sparse precision matrix #89
Description
After much digging and testing, it appears you are currently not supporting sparse matrix constructions for AD. Is this correct? In particular, I am in need of a MvNormal distribution that is specified using a sparse precision matrix and will work with the autodiff in Turing. This is common for large spatial problems using Gaussian Markov random fields. I came to Turing for this because stan does not support sparse matrix operations. This would be a really valuable addition to your package if you can make it work.