October 2019
Intermediate to advanced
340 pages
8h 39m
English
In this recipe, we randomly generated a policy that was composed of 16 actions for the 16 states. Keep in mind that in FrozenLake, the movement direction is only partially dependent on the chosen action. This increases the uncertainty of control.
After running the code in Step 4, you will see a 4 * 4 matrix as follows, representing the frozen lake and the tile (state 0) where the agent stands:

After running the lines of code in Step 5, you will see the resulting grid as follows, where the agent moves down to state 4:

An episode ...
Read now
Unlock full access