search

malllabiisc / CompGCN

ICLR 2020: Composition-Based Multi-Relational Graph Convolutional Networks
Apache License 2.0
597 stars 107 forks source link

What does gamma (margin) mean in your hyper parameter? #17

Closed ZifengDing closed 4 years ago

ZifengDing commented 4 years ago

Hello, Thanks for the great work. I am taking your work as my baseline. I notice that you have a parameter called gamma, which is "margin" in your description. As I understand, you did not include learning rate decay in your code. In the function "fit" of "run.py", what you did is just to stop training after continuous 25 epochs without improving MRR. So what exactly does this "margin" mean here?

Best Regards, Zifeng

soumyasanyal commented 4 years ago

Hi Zifeng,

The margin gamma is a specific parameter of the CompGCN_TransE model (check this line). You can find more details about the margin-based loss in the original TransE paper. We've also discussed the need to decrease gamma in this issue.

ZifengDing commented 4 years ago

Cool, now I get it. I was not taking TransE as the score function, so I missed this part. Thanks a lot for your quick response and help!