hexiangnan
commented
4 years ago The NDCG scores reported in the paper are based on a wrong implementation of NDCG. We have fixed it in the github but haven't revised the paper yet. We will revise the paper soon.
In the readme file, we have provided the correct NDCG scores.
On Fri, Apr 3, 2020 at 11:56 PM 刘昊东 notifications@github.com wrote:
Hi, thanks for your work. I used the hyper parameters you provide in the examples to run your code, though I can obtain satisfactory results about Recall on both three datasets, but the results about NDCG are far smaller than what you report in the paper. I am wondering what is suitable hyper parameters to obtain the result reported in your paper. Thanks.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kuandeng/LightGCN/issues/1, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAI3DDVVGXMU6A76FKFCEW3RKYBJTANCNFSM4L4GKGNQ .
-- Xiangnan HE (Dr) :: Professor :: Vice Dean, School of Data Science :: University of Science and Technology of China (USTC) :: 443 Huangshan Road, Hefei, China 230027 :: (0551) 63607236 (DID) :: hexn@ustc.edu.cn (E) :: http://staff.ustc.edu.cn/~hexn/ (W)
LiuHDme
Hi, thanks for your work. I used the hyper parameters you provide in the examples to run your code, though I can obtain satisfactory results about Recall on both three datasets, but the results about NDCG are far smaller than what you report in the paper. I am wondering what is suitable hyper parameters to obtain the result reported in your paper. Thanks.