Currently we cannot simulate the full DSA2000 run, which prevents us from seeing its impact on the results.
The main reason is that we run out of memory. Should explore packages for sampling on the fly without constructing the cholesky. We only need to be able to compute k(x,y) to compute the components of the cholesky matrix, and then this is simply a matrix multiplication with normally distributed variables, hence we should be able to do this without constructing the cholesky.
Currently we cannot simulate the full DSA2000 run, which prevents us from seeing its impact on the results.
The main reason is that we run out of memory. Should explore packages for sampling on the fly without constructing the cholesky. We only need to be able to compute
k(x,y)
to compute the components of the cholesky matrix, and then this is simply a matrix multiplication with normally distributed variables, hence we should be able to do this without constructing the cholesky.