Simulate and Compare Multi-Armed Bandit Strategies

Multi-Armed Bandit Simulation: Epsilon-Greedy, UCB1, and Thompson Sampling Introduction A multi-armed bandit (MAB) problem models repeated decision-making under uncertainty where an agent must choose among several actions (arms), each yielding a random reward drawn from an unknown distribution. The goal is to maximize cumulative reward over a finite horizon while learning the reward distributions. The core tension is exploration (trying less-known arms to gain information)…