search

MyTHWN / MTER

Explainable Recommendation via Multi-Task Learning in Opinionated Text Data
https://dl.acm.org/citation.cfm?id=3210010
39 stars 12 forks source link

How do you use Validation data when training MTER? #5

Open lthoang opened 5 years ago

lthoang commented 5 years ago

Hi Nan Wang,

As described in your paper, the split ratio for train/validation/test is 80/10/10. Is validation data only used for optimal hyper parameters search for baseline models? For MTER, do you use validation set for model selection purpose?

Thanks, Hoang

MyTHWN commented 5 years ago

Yes, the validation data is used to tune hyper-parameters for all models.

lthoang commented 5 years ago

Alright. I have a follow up question. What is the criteria that you used to select the best model for MTER? Best loss or best NDCG on validation set?

MyTHWN commented 5 years ago

Best NDCG