search

Thinklab-SJTU / GENN-Astar

Implementation of our CVPR 2021 paper "Combinatorial Learning of Graph Edit Distance via Dynamic Embedding".
Other
26 stars 3 forks source link

A question about precision_at_k in LINUX dataset #3

Open QAeQ opened 2 years ago

QAeQ commented 2 years ago

Hi, thanks for the excellent work. The precision_at_k in LINUX dataset is : image It seems that in AIDS dataset, the p@k of GENN is comparable with other methods, but in LINUX dataset, the p@k is 0.428, and it is far below than other methods like GraphSim or SimGNN, which is 0.992 and 0.942. Could you please explain why p@k is normal in AIDS dataset，but seems different in LINUX dataset? Thank you very much!

rogerwwww commented 2 years ago

Hi there,

Thanks for your interest!

GENN's accuracy is expected to be inferior to other methods because GENN is viewed as the simplified version of SimGNN/GraphSim. We care more about its efficiency so we remove some bells and whistles in the network.