You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
** Is there documentation missing? **
I think that the VariationalELBO mll computes the ELBO downscaled by the number of data points and this could be mentioned in the documentation. This can introduce some scaling issues when using the AddedLossTerm class with other losses that do not include this by default.
The text was updated successfully, but these errors were encountered:
tgcsaba
changed the title
[Docs] Describe in documentation of Variational ELBO normalized by num_Data
[Docs] Describe in documentation of Variational ELBO normalized by num_data
Dec 15, 2023
馃摎 Documentation/Examples
Hi, I am new to GPyTorch as I have just pivoted to using torch for my projects and enjoying the package. I am working on a project which includes a variational treatment of kernel hyperparameters in a similar spirit to Improving the Gaussian Process Sparse Spectrum Approximation by Representing Uncertainty in Frequency Inputs.
** Is there documentation missing? **
I think that the VariationalELBO mll computes the ELBO downscaled by the number of data points and this could be mentioned in the documentation. This can introduce some scaling issues when using the
AddedLossTerm
class with other losses that do not include this by default.The text was updated successfully, but these errors were encountered: