Steps: [ install <install> | jax <prereq_jax> | haiku <prereq_haiku> | q-learning | dqn <second_agent> | ppo <third_agent> | next_steps <next_steps> ]
In this first reinforcement learning example we'll solve a simple grid world environment.
Our agent starts at the top left cell, labeled S. The goal of our agent is to find its way to the bottom right cell, labeled G. The cells labeled H are holes, which the agent must learn to avoid.
In this example, we'll implement a simple value-based agent, which we update using the q-learning </examples/stubs/qlearning> algorithm.
To run this, either hit the Google Colab button or download and run the script on your local machine.
qlearning.py </examples/frozen_lake/qlearning.py>
/examples/frozen_lake/qlearning.py