Let’s get started with a simple project of estimating the value of π using the Monte Carlo method, which is the core of model-free reinforcement learning algorithms.
The multi-armed bandit problem is one of the simplest reinforcement learning problems. It is best described as a slot machine with multiple levers (arms), and each lever has a different payout and payout probability. Our goal is to discover the best lever with the maximum return so that we can keep choosing it afterward. Let’s start with a simple multi-armed bandit problem in which the payout and payout probability is fixed for each arm. After creating the environment, we will solve it using the random policy algorithm.
