How do you improve learning curve when working with Reinforcement Learning?

[evan_4gx3]

I am trying to improve my approach to Reinforcement Learning and could use some practical advice.

I am particularly interested in learning curve and data quality. I would like to compare approaches that people have actually used rather than just theoretical suggestions.

What would you tell someone trying this for the first time?