/pokemonai

Reinforcement learning approach to playing competitive Pokémon.

Primary LanguagePythonMIT LicenseMIT

Pokémon AI

Goal

Reinforcement learning approach to playing competitive Pokémon. Battling is done on Smogon's Pokémon Showdown! battle simulator.

Ideas to Explore

  • Batch RL (experience replay/fitted Q iteration)
  • Online TD Learning
  • PG methods (DDPG)
  • Monte Carlo Tree Search (with simplfied assumptions)
  • Imitation learning with Smogon leaderboard