Why test sample number is much more than train / validation sample?

[zhiqiangzhongddu]

I saw you assign train/val/test follow this way:
idx_train = range(140)
idx_val = range(200, 500)
idx_test = range(500, 1500)

I don't understand why you assign test sample number much more than train and val?

[tkipf]

To get a more reliable estimate of the test error :-)

…

On Thu 6. Sep 2018 at 13:58 Zhiqiang ZHONG ***@***.***> wrote: I saw you assign train/val/test follow this way: idx_train = range(140) idx_val = range(200, 500) idx_test = range(500, 1500) I don't understand why you assign test sample number much more than train and val? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#17>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHAcYDGUVQq-6xCD9EoMx9W5B2bkT3c7ks5uYRvogaJpZM4Wc4Xr> .

[zhiqiangzhongddu]

I see, thank you for the response.

[IreneZihuiLi]

Are these ids true ids in the cora cites dataset?
Or they are just from 0 to 1500.
Also, why id 140-200 were not chosen?

[tkipf]

These IDs don't reflect original paper IDs in Cora. We only select 140 labels (20 per class) to match the size of the splits used in our paper. No particularly striking reason for omitting IDs 140-200 apart from getting some "nicer" round numbers ;)

…

On Wed, Apr 3, 2019 at 2:59 AM IreneLi ***@***.***> wrote: Are these ids true ids in the cora cites dataset? Or they are just from 0 to 1500. Also, why id 140-200 were not chosen? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#17 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHAcYGz2PNhBB6N_XVVwOqdwCYa1rvdSks5vdAr1gaJpZM4Wc4Xr> .

[IreneZihuiLi]

Thanks!