search

valeoai / ADVENT

Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation
https://arxiv.org/abs/1811.12833
Other
382 stars 70 forks source link

About the entropy minimization #1

Closed Solacex closed 5 years ago

Solacex commented 5 years ago

Hello~ I notice that the log you are using in the entropy minimization is log2() rather than log(), have you tried torch.log() ?

tuanhungvu commented 5 years ago

Hello @Solacex , we haven't tried other bases. Using another base will simply scale the gradients by a constant scalar, which is no different from scaling the learning rate of entropy loss. T-H

Solacex commented 5 years ago

Hello @Solacex , we haven't tried other bases. Using another base will simply scale the gradients by a constant scalar, which is no difference from scaling the learning rate of entropy loss. T-H

Thank you for your prompt reply~

soans1994 commented 8 months ago

@tuanhungvu @Solacex

In the paper there are 2 methods, direct entropy minimization and adversarial training. Where is the first direct entropy minimization training code? There is only adversarial entropy minimization here?

thank you