import numpy as np

from collections import defaultdict

def epsilon_greedy(env, state, q, e = 0.95):

pi = defaultdict(lambda : 0)

actions = env.A(state)

best_action_idx = 0

for idx, action in enumerate(actions):

if(q[(state,action)] > q[(state,actions[best_action_idx])]):

best_action_idx = idx

for idx, action in enumerate(actions):

if(idx == best_action_idx):

pi[action] = 0.95

else:

pi[action] = (1-e)/(len(actions)-1)

return pi

def greedy(env, state, q):

pi = defaultdict(lambda : 0)

actions = env.A(state)

best_action_idx = 0

for idx, action in enumerate(actions):

if(q[(state,action)] > q[(state,actions[best_action_idx])]):

best_action_idx = idx

for idx, action in enumerate(actions):

if(idx == best_action_idx):

pi[action] = 1

else:

pi[action] = 0

return pi

def random(env, state, q):

pi = defaultdict(lambda : 0)

actions = env.A(state)

for action in actions:

pi[action] = 1/len(actions)

return pi