DDAN中参数p
#427
Replies: 1 comment 1 reply
-
你好,这是贺雯丽的自动回复,您的信息已经收到。祝身体健康,万事如意! 贺雯丽
|
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
您好,王研究员。在拜读您的文章“Transfer Learning with Dynamic Distribution Adaptation”时,您在最后进行参数敏感性分析,提到DDAN中的一个必要参数p,但我在您所写的相关的算法DeepMEDA中没有看到参数p出现在何处。


你在文章DDAN部分写道,模型的损失如下图。参数p体现在最后的拉普拉斯正则化项中。
但在算法中,总的损失计算如下图。
在算法计算损伤时,并没有体现出拉普拉斯正则化项,不知道是不是体现在weight中了。如果体现在weight中,那么参数p又体现在哪里呢?
可能是我对于算法还没有研究太深,问题可能比较浅显,希望您不吝赐教。
Beta Was this translation helpful? Give feedback.
All reactions