Forever
Results
3
comments of
Forever
same problem. sad.
关于Paper还有一个问题,论文里提到不直接匹配teacher和student的feature maps As for distribution matching, it is not a good choice to directly match the samples from it, since it ignores the sample density in the space. 可是从NST损失函数来看,也是在将两个分布之间的差异减小,不知这有什么区别。 这是论文中最疑惑的部分,希望得到指教,谢谢。
One more question 请问eq3中的第一项cross entropy loss在哪里?似乎只有kd和mmd loss