tadpole / dhne Goto Github PK

Hi, thank you for your work!
I have a question as stated in the title. For example. there is a hyper-edge "A2, U2, L1" in the test dataset, in the training process, you remove the hyper-edge, and then when predicting, you predict whether the hyper-edge exists, or to predict whether there is an edge between 'A2' and 'U2' (or 'A2' and 'L1' ...)?
Furthermore, is this method only suitable for the situation that there are links between any two vertices in the hyper-edge? Considering this situation that ‘there are edges between A1 and U1, U1 and L1, but there is no edge between A1 and L1’, is this method still suitable?

Looking forward to your reply! Thank you so much!

Hi!

Thanks for your good work!

I am not familiar with Tensorflow, so I want a pytorch version of DHNE, and the evaluation of the proposed DHNE.

What's more, other baselines are not found here.

Could you please share the corresponding code written with pytorch version and the code of evaluation of the performance of all methods used in this work?

非常感谢您的工作！
您好，我刚接触python，我想问一下这个程序的输出部分的代码语句在哪里呢？我没有找到。

如题，代码中默认的负采样比例是1:5，如果改成1:1，跑出来的AUC也只在0.6左右浮动，ACC表现也不太好

您好，输入的三元组是怎么抽的，是把网络中所有的三元组都抽出来，还是像random_walk一样抽样，可以重复，且数量由超参决定

Windows 10 Pro 1803 
Python3.6 
numpy==1.14.5

I've created a pull request #2 to fix the problem