Non Stationary GP Regression

Questions
3

Hi,

Thanks for your answer @bwengals

First of all, the modification didn’t work because Gibbs covariance function requires a callable function.

I tried making other changes like replacing lengthscale with the gp_l and variance with gp_sig but then the error would be 'Latent' object has no attribute 'ndim' which is occuring due to the variance product in cov_obs.

I also have two questions:

  1. Why are we taking outer product of covariance?

  2. Why do we need pm.gp.Latent? I understand there is no additive noise but then in the documentation, it is used when the data is not normally distributed.

Thanks,
Rohan