Hands-On Reinforcement Learning with Python

Solving the taxi problem using SARSA

Now we will solve the same taxi problem using SARSA:

import gymimport randomenv = gym.make('Taxi-v1')

Also, we will initialize the learning rate, gamma, and epsilon. Q table has a dictionary:

alpha = 0.85gamma = 0.90epsilon = 0.8Q = {}for s in range(env.observation_space.n):    for a in range(env.action_space.n):        Q[(s,a)] = 0.0

As usual, we define an epsilon_greedy policy for exploration:

def epsilon_greedy(state, epsilon):    if random.uniform(0,1) < epsilon:        return env.action_space.sample()    else:        return max(list(range(env.action_space.n)), key = lambda x: Q[(state,x)])

Now, the actual SARSA algorithm comes in:

for i in range(4000):        #We store cumulative reward of each episodes in r    r = 0     #Then for every iterations, ...