[orasis]

Reinforcement learning without simulation is not feasible for complex problems.

If you can’t simulate, stick to multi-armed bandits.