How to implement a reinforcement learning model?

Question

I have some question about reinforcement learning. I have studied all mathematical background of reinforcement learning but I am not able to code this model.

I am so confused about defining an environment. Can we use OpenAI gym for all kind of problems? It does not make sense to me. For example playing a game is different than stock prediction. But how come I see in both cases some people are using OpenAI gym?

Also I found the following list. What are thesis exactly? are they a frame work to implement a reinforcement learning model?

Could you please give a link, source, video , course or whatever that can make it clear for me.

Answer 1

OpenAI Gym is a collection of environments. If you studied the mathematical background of RL, then the Gym is simply a collection of MDPs. It is then up to you to hook an RL algorithm to those MDPs, and do actual learning.

That list you found is probably a bit old, as rllab has been renamed garage recently. I do not know most of them, but I would say they're libraries of RL algorithms already implemented (at least that's what rllab and OpenAI baselines are). You could look at their code to see how they implement the different RL algorithms.