Hi devlopers,
thanks for the great tool. I have one question with the fundamentals of the model architecture.
in TxGNN.py -> For finetuning, why is FullNegSampler used? because when you go to model in DistMult -> you just select specific edge types in edge_train.
Also, wouldn't it make more sense if you select the neg as contraindication and pos as indication?
Not, sure if that's what's happening in the model, difficult to interpret.
Thanks and Regards,
Onkar
Hi devlopers,
thanks for the great tool. I have one question with the fundamentals of the model architecture.
in TxGNN.py -> For finetuning, why is FullNegSampler used? because when you go to model in DistMult -> you just select specific edge types in edge_train.
Also, wouldn't it make more sense if you select the neg as contraindication and pos as indication?
Not, sure if that's what's happening in the model, difficult to interpret.
Thanks and Regards,
Onkar