search

Zhongdao / gcn_clustering

Code for CVPR'19 paper Linkage-based Face Clustering via GCN
MIT License
360 stars 86 forks source link

How to generate "attention aggregation" feature? #5

Closed jxyecn closed 5 years ago

jxyecn commented 5 years ago

Hello, I saw that you came up with a novel feature aggregation method "Attention Aggregation"(from 3.3), you said that the element in G is generate by a 2-layer MLP model. How to train this matrix, can you provide more detailed information? Thank you. (I didn't find relative source code, did I missed something?)

jxyecn commented 5 years ago

While reading your code, I find that "adjust matrix" A is reinitialized every batch. But from your paper, I understand that this A should be learned end to end. Is there anything wrong?

Zhongdao commented 5 years ago

This code contains only Mean Aggregation operation. Since the performance gain of Attention Aggregation is small, we do not use it in practice. You can check the paper Graph attention networks and their code for the implementation of Attention Aggregation.

jxyecn commented 5 years ago

OK, thanks for your reply!