hungpthanh/GRU4REC-pytorch

Difference from original paper: regularization term of TOP1 Loss

Closed this issue · 1 comments

bilzard commented

Hi.
Thank you for sharing this code.

I found difference from original paper in regularization term of TOP1 Loss.
According to the original paper, the regularization term should be calculated over only negative samples1.
However, this repo calculates it over all samples2.
It might not be significant difference, but I just pointed it out.

Footnotes

  1. https://github.com/hidasib/GRU4Rec/blob/master/gru4rec.py#L236 ↩

  2. https://github.com/hungthanhpham94/GRU4REC-pytorch/blob/master/lib/lossfunction.py#L89 ↩

hungpthanh commented

@bilzard Thank you for your comments!