MarginalLogLikelihood v/s LogMarginalLikelihood

[patel-zeel]

In the early stage of learning GPs, I was always confused between likelihood and marginal likelihood. For a certain period, I was thinking that training GPs is MLE because at some places I found people using term likelihood as a proxy for marginal likelihood. To make the difference very clear, isn't LogMarginalLikelihood more explicit than MarginalLogLikelihood? [first line of the answer]

[gpleiss]

To make the difference very clear, isn't LogMarginalLikelihood more explicit than MarginalLogLikelihood?

Yes, I do think this is the correct name. Changing things in GPyTorch would be possible, but would require appropriately deprecating the current naming convention.

[patel-zeel]

Thank you for the answer, @gpleiss. So, how would you suggest going about it?