A question about POG paper

[WxxShirley]

Hi,
I have just read your paper in KDD 19. It is really a nice work!
I am confused about the training process. In the POG framework, are the parameters in transition layers(both in Per and Gen ) are sharable with FOM models' transition layer parameters?
Or they are all learned during the training process with the proposed loss function?

Thanks!

[wenyuer]

Hi Xixi, Thank you for your interest. The parameters in POG's Gen network are initialized as the pre-trained FOM, so as the transition layers.  And also they can be learned during the training process, if your dataset is large enough.  Best Regards Wen Chen (陈雯)  

…

------------------&nbsp;原始邮件&nbsp;------------------ 发件人:&nbsp;"Xixi Wu"<notifications@github.com&gt;; 发送时间:&nbsp;2019年12月8日(星期天) 下午3:40 收件人:&nbsp;"wenyuer/POG"<POG@noreply.github.com&gt;; 抄送:&nbsp;"Subscribed"<subscribed@noreply.github.com&gt;; 主题:&nbsp;[wenyuer/POG] A question about POG paper (#1) Hi, I have just read your paper in KDD 19. It is really a nice work! I am confused about the training process. In the POG framework, are the parameters in transition layers(both in Per and Gen ) are sharable with FOM models' transition layer parameters? Or they are all learned during the training process with the proposed loss function? Thanks! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[WxxShirley]

I got it.
Thanks!

[hyyuan123]

Hi,Is there a public code for pog papers?