Using deep reinforcement learning algorithms to learn how to battle in the game Pokémon.
Q-Learning, Asynchronous Advantage Actor-Critic, and NeuroEvolution of Augmenting Topologies approaches were used to solve steadily more complex environments where the agent is given one or more Pokémon and expected to defeat the opposing trainers.
Please see the pdf titled "Pokémon Deep Reinforcment Learning" above for more details.