Hands-On Reinforcement Learning with Python

The epsilon-greedy policy

We have already learned a lot about the epsilon-greedy policy. In the epsilon-greedy policy, either we select the best arm with a probability 1-epsilon or we select the arms at random with a probability epsilon:

Now we will see how to select the best arm using the epsilon-greedy policy:

1. First, let us initialize all variables:

# number of rounds (iterations)num_rounds = 20000# Count of number of times an arm was pulledcount = np.zeros(10)# Sum of rewards of each armsum_rewards = np.zeros(10)# Q value which is the average rewardQ = np.zeros(10)

2. Now we define our epsilon_greedy function:

def epsilon_greedy(epsilon): ...