backgom2357/Recommender_system_via_deep_RL

How to improve precision?

Opened this issue · 2 comments

Dear author, may I ask why the precision of this model is only about 0.4? Is it possible to improve it? Where did you get the other two files:“actor_50000.h5” “critic_50000.h5”

Dear author, may I ask why the precision of this model is only about 0.4? Is it possible to improve it? Where did you get the other two files:“actor_50000.h5” “critic_50000.h5”

Hi, Panda0907! The main reason is, i think, the 'session'. The simulator from paper use the session, so that agent may search next item in small range and learn short-term pattern of the user. As i use whole data in a single training, the performance may be lower.

The .h5 files are not given, but you can get when you run train.py

I have run train.py but file “actor_50000.h5” “critic_50000.h5” cannot be obtained. please help me