Julia implementation of DeepMind's Deep Q-Learning algorithm as described in Playing Atari with Deep Reinforcement Learning. This code only implements the base algorithm. It does not include the code for a convolutional network. However, this can be easily added using Mocha.jl. In lieu of this it uses a simpler single layer neural network. Information on the original RecurrenJS DQN implementation can be found here

note: This library has been tested on various learning tasks and seems to be functioning correctly, but is not yet ready for public consumption.

using DeepQLearning

... coming soon ... I hope :)

##Dependencies This library requires NNGraph.jl.

##Credits This library draws on the work of Andrej Karpathy

MIT