dropout in validation/evaluation

[Atomu2014]

非常感谢分享这份代码,
能否解释一下第134和143行中, dropout_keep这个参数在init_graph的时候直接传入tf.nn.dropout

[Atomu2014]

请问在validation的时候是否也用了drop out，这是合理情况吗？在test的时候是否会使用dropout？

[hexiangnan]

只有training时使用dropout。validation 和 testing时必须disable dropout.

…

On Thu, Feb 15, 2018 at 2:44 PM, KevinKune ***@***.***> wrote: 请问在validation的时候是否也用了drop out，这是合理情况吗？在test的时候是否会使用dropout？ — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABGxjhT99zz2sz_ARYG_9IJTalTZ7Ho7ks5tU9JdgaJpZM4R1huB> .

-- Best Regards, Xiangnan He

[Atomu2014]

貌似training和validation共用一套计算图，而且dropout在计算图中是写死的，没有发现training flag之类的分支

[hexiangnan]

train_phase 这个flag用户区分是在training还是testing/validation.

…

On Mon, Feb 26, 2018 at 11:02 AM, KevinKune ***@***.***> wrote: 貌似training和validation共用一套计算图，而且dropout在计算图中是写死的，没有发现training flag之类的分支 — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#8 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABGxjhlsyWmD7k2I0XFNC80Ae6vapAw7ks5tYh7IgaJpZM4R1huB> .

-- Best Regards, Xiangnan He

[Atomu2014]

感谢告知！