a-bone1 / esam Goto Github PK

Hi~
I have carefully read your code in test/test_movielen.py and have run for it.
But I fail to found that any features bot not item_id/movie_id have utilizing into your network.
Could you share how you incorporate other feature, e.g. genres of the movie into your training?

Hi~
What a great work.
Would you share the dataset of CIKM Cup 2016 Personalized E-Commerce Search. The link in the paper was failed.

Has any kind soul written a PyTorch version of this code? Could someone please share it with me?

in ./ESAM/test/center_loss.py，the function
def get_center_loss(features, labels, num_classes):
...
return (loss_0 + loss_1)/(210+22)

why divide by (210+22) ？
why is (210+22)？

In the file test_movielens.py, the class NeuMF_my_tail() is imported. But it seems only the cross-entropy loss and regularization loss are used.
But as the paper put it, the loss contains Domain Adaptation Loss, Center-wise Loss and Pseudo Label Loss.
Have I missed anything?