In a time series model with regressors, I would like to model the trend with a Gaussian process. I see that I can describe the model either with a Marginal Gaussian process (with y=None and a small noise) or a Latent Gaussian process. Both descriptions give the same results, though the number of div…

Marginal Gaussian with y=None vs Latent Gaussian

tbosois May 12, 2025, 1:28am 2

If I’m not mistaken, the Latent GP is a 50 dimensional vector whereas the marginal GP is a 4 dimensional vector. Easy to see why then it is more prone to errors.

Topic		Replies	Views
GP and Bayesian linear regression in PyMC Questions	8	1434	August 18, 2018
Gaussian Processes. Latent vs Marginal gaussian_process , modeling	2	518	October 24, 2023
How do you formulate a `Latent` GP that is equivalent to the `Marginal` implementation? Questions	5	503	May 26, 2021
GP marginal likelihood with white noise/Latent GP Questions	3	496	June 3, 2019
Latent gaussian process prediction problem Questions gaussian_process	3	1097	January 25, 2022

Marginal Gaussian with y=None vs Latent Gaussian

Related topics