Is it possible to get prior predictive samples for Gaussian process models? I’d like something like this (MWE taken from the docs):
x = np.linspace(0,1,50) y = 0.5 *x + np.random.randn(50) * 0.01 with pm.Model() as gp_model: cov_func = pm.gp.cov.ExpQuad(1, ls=0.1) gp = pm.gp.Marginal(cov_func=cov_func) sigma = pm.HalfCauchy("sigma", beta=5) y_ = gp.marginal_likelihood("y", X=x[:,None], y=y, noise=sigma) y_star = gp.conditional("y_star", x[:,None]) with gp_model: prior = pm.sample_prior_predictive()
to create prior predictive samples for y_star? However atm this always returns an array of NaNs.