文章目录
class_loss
分类交叉熵损失_论文公式(2)
ring_loss
Ring-style Normalization_论文公式(8)
mutual_information
_论文公式(5/6/7)
confusion_loss
L
e
n
t
L_{ent}
Lent_论文公式(3)
alignment_loss
FD
就是论文中的DI
Specifically, we leverage a domain identifier D I DI DI, which takes the disentangled feature ( f d i f_{di} fdi or f d s f_{ds} fds ) as input and outputs the domain label l f l_f lf(source or target).
Reconstruction
重构损失_论文公式(1)
与论文对比
论文
代码流程