/
sarsa.py
79 lines (65 loc) · 2.32 KB
/
sarsa.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
import numpy as np
from snake import SnakeEnv, ModelFreeAgent, TableAgent, eval_game
from policy_iter import PolicyIteration
from monte_carlo import MonteCarlo, timer
class SARSA(object):
def __init__(self, epsilon=0.0):
self.epsilon = epsilon
# sarsa的策略评估
def sarsa_eval(self, agent, env):
state = env.reset()
prev_state = -1
prev_act = -1
while True:
act = agent.play(state, self.epsilon)
next_state, reward, terminate, _ = env.step(act)
if prev_act != -1:
# SARSA的迭代公式
return_val = reward + agent.gamma * (0 if terminate else agent.value_q[state][act])
agent.value_n[prev_state][prev_act] += 1
agent.value_q[prev_state][prev_act] += (return_val - agent.value_q[prev_state][prev_act]) / agent.value_n[prev_state][prev_act]
prev_act = act
prev_state = state
state = next_state
if terminate:
break
def policy_improve(self, agent):
new_policy = np.zeros_like(agent.pi)
for i in range(1, agent.s_len):
new_policy[i] = np.argmax(agent.value_q[i, :])
if np.all(np.equal(new_policy, agent.pi)):
return False
else:
agent.pi = new_policy
return True
# sarsa
def sarsa(self, agent, env):
for i in range(10):
for j in range(2000):
self.sarsa_eval(agent, env)
self.policy_improve(agent)
def monte_carlo_demo():
np.random.seed(101)
env = SnakeEnv(10, [3, 6])
agent = ModelFreeAgent(env)
mc = MonteCarlo(0.5)
with timer('Timer Monte Carlo Iter'):
mc.monte_carlo_opt(agent, env)
print('return_pi={}'.format(eval_game(env, agent)))
print(agent.pi)
np.random.seed(101)
agent2 = TableAgent(env)
pi_algo = PolicyIteration()
with timer('Timer PolicyIter'):
pi_algo.policy_iteration(agent2)
print('return_pi={}'.format(eval_game(env, agent2)))
print(agent2.pi)
np.random.seed(101)
agent3 = ModelFreeAgent(env)
mc = SARSA(0.5)
with timer('Timer Monte Carlo Iter'):
mc.sarsa(agent3, env)
print('return_pi={}'.format(eval_game(env, agent3)))
print(agent3.pi)
if __name__ == '__main__':
monte_carlo_demo()