AFM训练中出现nan

[Atomu2014]

AFM训练起来很容易nan，请问您遇到过这种情况吗？对于调参数有什么建议？哪些参数比较敏感？

[hexiangnan]

nan应该是因为我们自己用exp函数实现了softmax, 导致容易overflow。换用tensorflow实现的softmax应该就不会nan了。

…

On Sat, Feb 10, 2018 at 11:48 PM, KevinKune ***@***.***> wrote: AFM训练起来很容易nan，请问您遇到过这种情况吗？对于调参数有什么建议？哪些参数比较敏感？ — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#9>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABGxjuXIl51KSyJ121ucnDpYetMQ64Wiks5tTboxgaJpZM4SA-Kv> .

-- Best Regards, Xiangnan He

[Atomu2014]

感谢回答，最近发现调节softmax的温度会比较容易训练，在我的数据集上调节温度比attention_lambda更有效，能显著控制过拟合，建议代码中增加调节温度的参数

[hexiangnan]

谢谢告知！

…

On Thu, Feb 15, 2018 at 2:50 PM, KevinKune ***@***.***> wrote: 感谢回答，最近发现调节softmax的温度会比较容易训练，在我的数据集上调节温度比attention_lambda更有效，能显著控制过拟合， 建议代码中增加调节温度的参数 — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#9 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABGxjgrFgN4ZzcHQz8khApezHDNIjFkmks5tU9PIgaJpZM4SA-Kv> .

-- Best Regards, Xiangnan He