Teaching an AI Agent to play the game FrozenLake using Reinforcement Learning - UE4
Fun weekend practice on reinforcement learning by reproducing the FrozenLake demo from the OpenAI Gym library in Unreal Engine 4 then training an agent to navigate its way to the goal using Markov Decision Processes and value iteration with Reinforcement Learning.
FrozenLake environment is stochastic, which makes it even difficult for the agent to reach the Goal without falling in a hole.
The agent will practice for 10,000 episodes then it will play on its own for 5 rounds.
let's see how the agent it's going to perform. 😁