My Solution

[Vvkmnn]

Thanks for the tutorial/video. This was fun. May have gone a bit overboard. xD

Wish I'd made it smarter, might play with the epsilon idea in the future. I've reinforced my hunger for reinforcement learning!

Elements of my solution:

• random_start() start anywhere on the board.
• difficulty parameter that scales up the number of walls.
• create_reds(),create_greens() and create_walls() that add more special squares to help and hinder the agent.
• (x,y) scaling for larger boards
• max_q() changes for more randomized/Q-sensitive agent decision making.
• Tons more documentation (for my own learning)

[baob]

Would have been great to see some of those changes in separate commits.