[mlpack] Structuring mlpack RL

[Adithya Praveen]

Hey there! Adi here.
>From what I understand, right now DQN's and other methods in mlpack are implemented in only the test file for "q_learning", and are limited to the environments implemented in mlpack. If I'm not mistaken, these environments return state spaces / observations that are limited to vectors. 
1. What I'm trying to say is, I'd like to implement environments that return image tensors as the state, maybe something like pong. However, I'm not sure if that is an avenue you guys would like to pursue. Do we plan on,
a. adding more environments to mlpack (if so, guidance on how to go about doing that would be very helpful for me) or,
b. use Marcus's api in the future (and maybe include tutorials in mlpack's documentation on how to use mlpack's RL agents in unison with the api). Because I feel the api hasn't got the exposure it deserves, and I would love to add documentation (maybe as an mlpack tutorial). Hopefully I'll be able to figure out how to use mlpack's RL method's with zoq's api.
Again, I would love to get your take on these options (or other options that I haven't accounted for :)).
2. I would also like to know if adding an "agents" folder with different RL algos, to the RL section makes sense? You know, so that mlpack user's could just create, say, a "DQN Agent" for an environment and try running it.
I know that this is a loaded email. However, I would love to work on this for gsoc as well, and knowing your take on all this would be invaluable information.
Cheers,Adi.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://knife.lugatgt.org/pipermail/mlpack/attachments/20200228/02720922/attachment.htm>