Investigate Validation Loss < Testing Loss

paul-tqh-nguyen / anime_recommendation_system_comparison

A comparison of matrix factorization vs a deep learning model for recommending animes.

0 stars 0 forks source link

Investigate Validation Loss < Testing Loss #22

Open paul-tqh-nguyen opened 4 years ago

paul-tqh-nguyen commented 4 years ago

This is usually a sign of overfitting. Investigate and let's see what's going on.

paul-tqh-nguyen commented 4 years ago

We might be overfitting because we're using embedding vectors with size > 100 and many users have only 100 reviews. This is a high priority issue to fix.