I started roughing out the sparse matrix functional spec
(also vectors and row vectors):
The stumbling block is going to be posterior analysis where
we need every parameter, transformed parameter and generated
quantity to have the same shape so we have the same variables
in each iteration. But if we have to declare a sparse matrix
with the coordinates (row, column pairs) at which it is potentially
non-zero, it’s going to be very hard to do when you have operations
on sparse matrices like addition.
- Bob