An available reinforcement learning framework for reference

[flyers]

I found an available open source reinforcement learning framework here https://github.com/rllab/rllab We can take it as a reference when implementing some RL algorithms. Also, the relevant paper can be found here http://arxiv.org/abs/1604.06778

[sxjscience]

This is really helpful. We can replicate the RL algorithms mentioned in the paper and experiment parallel + distributed settings.

[sxjscience]

We need to compare the speed between Arena and other reinforcement learning toolboxes, e.g.

• [ ] Atari (https://github.com/Kaixhin/Atari)
• [ ] rllab (https://github.com/rllab/rllab)
• [ ] tflearn (https://github.com/tflearn/tflearn/blob/master/examples/reinforcement_learning/atari_1step_qlearning.py)

Following are some optimizations we can do

• [ ] Rewriting the replay memory using C++.
• [ ] Rewriting Reinforcement Learning related Operators using C++.