search

Lin-Yijie / Graph-Matching-Networks

PyTorch implementation of Graph Matching Networks, e.g., Graph Matching with Bi-level Noisy Correspondence (COMMON, ICCV 2023), Graph Matching Networks for Learning the Similarity of Graph Structured Objects (GMN, ICML 2019).
Other
282 stars 55 forks source link

Scaling of edge network with 0.1 #13

Open tsjain opened 3 months ago

tsjain commented 3 months ago

I did not see where the edge network weights are scaled by 0.1 in the code.

Lin-Yijie commented 3 months ago

Well, we do not implement the Glorot initialization with a scaling factor of 0.1. Instead, we directly use the default initialization of torch. If you figure out how to change this, you could provide a pull request.