search

aI-area / DVGSN

0 stars 1 forks source link

Want to ask whether this paper is published? #1

Open Xiefeng69 opened 2 years ago

Xiefeng69 commented 2 years ago

This is such an amazing work. I downloaded and read 《Dynamic Virtual Graph Significance Networks for Predicting Influenza》 on arXiv, and found the source code. The model proposed in the paper also gave me new ideas. But when I reproduced this model, I found a problem. You modeled all the time-point data at the beginning, so the subsequent predictions actually are meaningless, because you already know the future data. I think when calculating the significant of edges, the time-point in the validate set / test set should not be involved in the significant calculation of the whole graph. I don't think the design of this part is reasonable, maybe I understand it wrong.

But the idea is very good, can I quote this paper on arXiv in the future?

best wish.

aI-area commented 2 years ago

Sorry for responding so late...We have not noticed your comment. As we have mentioned in the paper, We have split the data into the training/validation/test groups from different time points.