descriptinc / melgan-neurips

GAN-based Mel-Spectrogram Inversion Network for Text-to-Speech Synthesis
MIT License
980 stars 214 forks source link

Some questions about your paper #31

Open XinMing0411 opened 4 years ago

XinMing0411 commented 4 years ago

The discriminator loss function your used in paper us is" min(0,1-D_k(x))+ min(0,1+D_k(G(s,z)))" ,but u said in your paper it is hinge loss ,and u used the Relu method in your code. So I think it should be " max(0,1-D_k(x))+ max(0,1+D_k(G(s,z)))". Why is that

Maybe I am wrong QAQ