search

evison / JRL

Code for paper: Joint Representation Learning for Top-N Recommendation with Heterogenous Information Sources
GNU General Public License v3.0
49 stars 20 forks source link

Experiment results in the spllited dataset. #4

Closed qq472220372 closed 5 years ago

qq472220372 commented 5 years ago

I set the params as the paper introducted. Category | net_struct | epoch | rate_emb_size | loss | LRS@10 | Recall@10 | Precision@10 | NDCG@10 | Hit@10 | LR | img_w | emb_size | grad_norm Beauty_CIKM2017 | default | 20 | 200 | 2080.8 | 0.01184 | 0.02115 | 0.00389 | 0.01315 | 0.04087 | 0.5 | default | 300 | 10 Can you give me some advices? Thank you so much!

0215Arthur commented 5 years ago

Hi~ I just try to get the splitted dataset you provided https://www.dropbox.com/s/th672ttebwxhsfx/CIKM2017.zip?dl=0 Howerver, the zip file failed to open and it showed that the file is wrong. Maybe the data is broken? Could you give me some advices of getting the splitted data? @evison

evison commented 5 years ago

@0215Arthur Thanks for inquiry. Some people also has similar problem. Usually it is because the machine used to unzip the file does not has sufficiently large memory, or because of the memory limit of commends if you are unzipping the file using a terminal. I would suggest you to unzip the file on a machine with larger memory, and use "jar xvf" command instead "unzip" command to unzip the file.

0215Arthur commented 5 years ago

@evison OMG ! Thank you very much! I will have a try.