Standalone implementations of reinforcement learning algorithms

[jean72human]

Hello,
I would like to know what you think about having some standalone implementations as functions that take in the environment and other parameters and return the trained policy.

Here an example of how this could look like for deep Q learning:

policy = deepq.learn(env, network=q_function_approximator, lr=learning_rate, epsilon=exploration_rate, buffer_size=buffer_size)

I think this would make it easier to quickly get started with deep reinforcement learning with Flux

[tejank10]

It'd be great to have a sister package, like what baselines is to OpenAI Gym, showing off RL algorithms with Gym.jl

[jean72human]

Should it be a seperate package or can it be included in Gym.jl?

[v-i-s-h]

I'm also really interested to see this happening. Making it a separate package will be useful, I think.
Otherwise, users who want to use Gym.jl alone will also have to download the dependencies of this.

Further, which framework you propose will be best suited for this: Knet, Flux or Tensorflow.jl?

[tejank10]

Making separate package is useful, for the same reasons mentioned by @v-i-s-h . Any of the framework can be used, but I'll be biased towards Flux ;)

[jean72human]

I think Flux would be good for that