CowRy

Results 4 issues of CowRy

https://github.com/qcymkxyc/RecSys/blob/6f8067f8b4ac40af7b7f6099aa5958deba8356e3/main/chapter2/lfm.py#L109 没有加入误差

https://github.com/qcymkxyc/RecSys/blob/6f8067f8b4ac40af7b7f6099aa5958deba8356e3/main/chapter2/lfm.py#L34 书中的负采样描述的是 对每个用户采样负样本时,要选取那些很热门,而用户却没有行为的物品。且书中的item_pool是一个列表,我的理解是不用去重,列表中重复的次数代表着物品的流行度,流行度大的更容易被采样到。但这样可能存在items列表过大... 所以可以考虑得到去重之后的物品列表后,记录每个物品的出现次数,与总个数相处得到频率,以此作为np.choice抽样的概率。

The author of hop-rec did not publish the source code, and I did not find any other version on the Internet。Could you please release the code of hop-rec? Thank you~

my environment ``` torch 1.1.0 torch-cluster 1.3.0 torch-geometric 1.2.1 torch-scatter 1.2.0 torch-sparse 0.4.0 torch-spline-conv 1.1.0 ``` log ``` python main.py --dataset=diginetica 2019-07-05 16:06:52,703 main.py[line:34] Namespace(batch_size=1024, dataset='diginetica', epoch=10, hidden_size=100, l2=1e-05, lr=0.001,...