I want to get reproducible results from GP.sample. This function could take an optional parameter rng of type numpy.random.Generator. Rather than call np.random.multivariate_normal, we could then call rng.multivariate_normal.
The same holds for GP.sample_conditional. These seem to be the only two functions in George that rely on numpy.random.
I want to get reproducible results from
GP.sample. This function could take an optional parameterrngof typenumpy.random.Generator. Rather than callnp.random.multivariate_normal, we could then callrng.multivariate_normal.The same holds for
GP.sample_conditional. These seem to be the only two functions in George that rely onnumpy.random.