Gymnasium Taxi 🚕

Overview

• Notebook has two algorithms DDNQ with Prioritization and traditional QTable
• This is for a school project and learning, nothing fancy
• It also creates videos to see the performance of the agent

Results

• We can see that DDNQ w Prio learns faster (less episodes) than QTable
• QTable aprox. requires 3500 episodes, rather than the DDQN requires 1500 to reach the same reward average
• On the other hand, QTable is quicker overall, since processing each iteration requires less resources (or CPU power)