Try run AttRec but the HR sorce is too low

cheungdaven / DeepRec

An Open-source Toolkit for Deep Learning based Recommendation with Tensorflow.

GNU General Public License v3.0

1.13k stars 294 forks source link

Closed molybdenumyz closed 5 years ago

molybdenumyz commented 5 years ago

HR@50:0.3753976670201485 MRR:0.053446402676701824 AUC:0.8771433443062608 92456 Epoch: 0543; (943, 100) how can I reproduce effect

cheungdaven commented 5 years ago

It seems that I put a wrong loss there, I will revise it, thank you for pointing out.

cheungdaven commented 5 years ago

https://github.com/cheungdaven/DeepRec/commit/a6c93234c876777d9fe45ee719eff72c41ae1c49 I have updated the loss function, It should reproduce the reported results now. If you have any other questions, just leave me a message here.

cheungdaven commented 5 years ago

The best results I can get now: HR@50:0.5312831389183457 MRR:0.0911185864781581 AUC:0.909033997832354

DebonairLi commented 5 years ago

The best results I can get now: HR@50:0.5312831389183457 MRR:0.0911185864781581 AUC:0.909033997832354

In your code, maybe there aren't any parts to calculate the best HR, MRR，we need to revise it,right?

cheungdaven commented 5 years ago

yeah, there are not. I will keep improving the repo. Thanks.