I’m training on RGCN with pytorch, I want to get the accuracy or MRR of each relation, is it possible?
I’m thinking of making several testing dataset, each of dataset consists of only one relation, and test them one by one by the trained and saved RGCN model, would this idea work?
I’m pretty new to computer science, an example would be very helpful.
If accuracy of each relation means training use each relation separately, the way you mention works.
If it is to measure the ratio each relation takes to overall model(trained using all relations), this doesn’t work. To achieve this, maybe check the final trained weight of each relation is more reasonable.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.