in trainer.py /gen_update(), i cannot understand "encoder is tuned, input is fixed" and "encoder is fixed, input is tuned"

NVlabs / DG-Net

:couple: Joint Discriminative and Generative Learning for Person Re-identification. CVPR'19 (Oral) :couple:

https://www.zdzheng.xyz/publication/Joint-di2019

Other

1.27k stars 230 forks source link

in trainer.py /gen_update(), i cannot understand "encoder is tuned, input is fixed" and "encoder is fixed, input is tuned" #67

Closed wdf19961118 closed 3 years ago

layumi commented 3 years ago

Hi @wdf19961118 It is a two-stage strategy. If you are familiar with EM (Expectation Maximization Algorithm), the mechanism is somehow similar. We can not update input and encoder at the same time, which may lead to the optimisation problems such as overfitting.