Skip to main content

Reinforcement Learning Agent Matrix

Created on September 30|Last edited on October 1
All the videos from one training run of a reinforcement learning agent in a grid world are plotted on two dimensions:
  • X axis is side effects: the more of a mess the agent made, the further to the right
  • Y axis is reward: the more the agent accomplished, the higher up
Hover over individual points to see the videos. The second panel is filtered to show only the successful agents (ones that complete the game level and win).


Maybe<(table)>
Maybe<(table)>