This is an attempt at faithfully recreating the set of DQN variants discussed in the paper "Rainbow: Combining Improvements in Deep Reinforcement Learning". I made this as part of my work getting up to speed in Deep RL under a grant from the Machine Intelligence Research Institute.
If all prerequisites are installed, can be run in the default configuration on Pong by typing "python rainbow.py".
Here's a graph of performance on Pong with all variants enabled: