search

1049451037 / GCN-Align

Code of the paper: Cross-lingual Knowledge Graph Alignment via Graph Convolutional Networks.
128 stars 27 forks source link

About the parameter settings #2

Closed yerui51 closed 5 years ago

yerui51 commented 5 years ago

I am very interested in your papers and want to reproduce your experiment. I would like to ask about the settings of your parameters， including the learning rate, number of negative samples, number of iterations of structure and attribute. Thank you very much. Eager for your reply!

1049451037 commented 5 years ago

Hi, thank you for your interest! The parameters are in the include/Config.py. The parameters here are fine-tuned afterwords, which performs better than the paper, the parameters of the experiments are demonstrated in the paper.

yerui51 commented 5 years ago

Hi, thanks for your reply. I have successfully run your code. When I read the code, I found a small error, in the get_sparse_tensor function, ind.append((sec, fir)), sec and fir，is their position reversed? . I would like to ask if my understanding is biased or your small mistake, thank you.

1049451037 commented 5 years ago

I re-checked it and I think it is right. In the paper we define that $a_ij$ is the effect of i-th entity to j-th entity. But here the entity matrix is left multiplied by the M matrix, which means $M_ij$ equals $a_ji$. Hope it helps!