What‘s the difference between GPM and GCNZ with few layer

yinboc / DGP

Rethinking Knowledge Graph Propagation for Zero-Shot Learning, in CVPR 2019

MIT License

320 stars 57 forks source link

What‘s the difference between GPM and GCNZ with few layer #4

Closed Wangt-CN closed 5 years ago

Wangt-CN commented 5 years ago

Hi,

It's a great work, but i have some confusions about GPM and GCNZ.

From the reported result in paper, GPM has already received a huge improvement compared to the SOTA method GCNZ, however the paper just said GPM only use single layer compared to GNCZ.

However in GNCZ, author's experiment indicate that addition layers bring the improvement results. So i'm very confused.

Can i ask more clear difference between GPM and GCNZ?

Thanks very much!

yinboc commented 5 years ago

Thanks for the question.

The main differences are: 1. number of hidden channels 2. training epochs Both contribute much improvement to GCNZ-1.

1: GPM: 2048 GNCZ-1: 512 GNCZ-6: 2048, 2048, 1024, 1024, 512

2: GPM 3000 epochs, GNCZ 300 epochs

Wangt-CN commented 5 years ago

Great and that's clear！Thank you for your reply!