Link prediction validation loss lower

Questions
#1

What could be the reasons for validation loss to be lower in link prediction task with max-margin loss? I do transductive learning with negative sampling and I keep getting lower validation loss - cant understand why.

Has anybody been in this situation and figured out why?

Could someone answer these questions please?
#2

Let’s discuss here since this is a duplicate question.

#3

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.