本文整理汇总了Python中agent.Agent.td_learning方法的典型用法代码示例。如果您正苦于以下问题：Python Agent.td_learning方法的具体用法？Python Agent.td_learning怎么用？Python Agent.td_learning使用的例子？那么恭喜您, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类agent.Agent的用法示例。
在下文中一共展示了Agent.td_learning方法的2个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: Environment
# 需要导入模块: from agent import Agent [as 别名]
# 或者: from agent.Agent import td_learning [as 别名]
import numpy as np
from utils import compute_mse, Trace
if __name__ == '__main__':
"""
test sarsa lambda algorithm
"""
""" the learning curve of mean-squared error
against episode number for lambda = 0 and lambda = 1
"""
env = Environment()
agent = Agent(env)
print ('the learning curve of mean-squared error against episode number for')
print ('lambda = 0')
agent.td_learning(10000, 0.0, True, trace = Trace.accumulating)
agent.reset()
print ('lambda = 1')
agent.td_learning(10000, 1.0, True, trace = Trace.accumulating)
agent.reset()
print ('The mean-squared error against lambda')
monte_carlo_iterations = 1000000
td_iterations = 10000
agent.monte_carlo_control(monte_carlo_iterations)
Q_monte_carlo = agent.Q
alphas = np.linspace(0,1,11)
mse_all = []示例2: Environment
# 需要导入模块: from agent import Agent [as 别名]
# 或者: from agent.Agent import td_learning [as 别名]
td_iterations = 10000
env = Environment()
agent = Agent(env)
agent.monte_carlo_control(monte_carlo_iterations)
Q_monte_carlo = agent.Q
alphas = np.linspace(0,1,11)
mse_all_acc = []
mse_all_replace = []
mse_all_dutch = []
avg_iters = 10
for alpha in alphas:
mse_current = 0
for i in range (0,avg_iters):
agent.reset()
agent.td_learning(td_iterations, alpha, trace = Trace.accumulating)
Q_tf = agent.Q
mse_current += compute_mse(Q_tf, Q_monte_carlo, True)
mse_all_acc.append(mse_current / avg_iters)
mse_current = 0
for i in range (0,avg_iters):
agent.reset()
agent.td_learning(td_iterations, alpha, trace = Trace.replacing)
Q_tf = agent.Q
mse_current += compute_mse(Q_tf, Q_monte_carlo, True)
mse_all_replace.append(mse_current / avg_iters)
mse_current = 0