Multidimensional gaussian process

Questions
4 
params = np.array([[1.0, -2.4, 3.6, 1.3]])

(one more pair of brackets)

which has 1 row and 4 columns. Also like I’d suggest a couple other things that might be helpful,

  • Use gp.Marginal, since the likelihood is MvNormal it’s conjugate to the GP. You’ll get a big speed up. If you wish to use gp.Latent with that likelihood, since your covariance for MvNormal is diagonal, you can use Normal instead which will be more efficient.
  • Set njobs=1 in the pm.sample(...) call. The matrix operations used by Theano here are multithreaded, so running multiple chains simultaneously bogs things down.
  • There should be no need to set start=model.test_point if everything is specified properly (should be gp_model.test_point I think).
1 Like