samet-akcay / ganomaly

GANomaly: Semi-Supervised Anomaly Detection via Adversarial Training
MIT License
853 stars 213 forks source link

why do forward_d() twice? #69

Open niuwagege opened 4 years ago

niuwagege commented 4 years ago

in backward_g(),call self.netd() to get features. self.err_g_adv = self.l_adv(self.netd(self.input)[1], self.netd(self.fake)[1]) why not use self.feat_real and self.feat_fake that you get from forward_d() directly