系列文章目录
强化学习
提示：写完文章后，目录可以自动生成，如何生成可参考右边的帮助文档
前言
强化学习（Reinforcement Learning, RL），又称再励学习、评价学习或增强学习，是机器学习的范式和方法论之一，用于描述和解决智能体（agent）在与环境的交互过程中通过学习策略以达成回报最大化或实现特定目标的问题 。
一、强化学习是什么？
强化学习是智能体（Agent）以“试错”的方式进行学习，通过与环境进行交互获得的奖赏指导行为，目标是使智能体获得最大的奖赏，强化学习不同于连接主义学习中的监督学习，主要表现在强化信号上，强化学习中由环境提供的强化信号是对产生动作的好坏作一种评价(通常为标量信号)，而不是告诉强化学习系统RLS(reinforcement learning system)如何去产生正确的动作。由于外部环境提供的信息很少，RLS必须靠自身的经历进行学习。通过这种方式，RLS在行动-评价的环境中获得知识，改进行动方案以适应环境。
理解：强化学习其实就是和人一样，一开始是什么都不懂的，所谓吃一堑长一智，他像一个新生的孩子，它在不断的试错过程中慢慢知道了做什么有奖励，做什么对得到奖励会有一定的价值，做什么会被打。在这个过程中不会像监督学习一样有个师傅带你，完全需要自己去摸索，就像修仙宗门一样，有背景的宗门弟子是继承掌门之位（监督），创立宗门的人是开山立派（强化），必须一步一个脚印去不断成长。
其实强化学习吸引我的就是因为它主要使用在游戏上，例如：
在 Flappy bird 这个游戏中，我们需要简单的点击操作来控制小鸟，躲过各种水管，飞的越远越好，因为飞的越远就能获得更高的积分奖励。
机器有一个玩家小鸟——Agent
需要控制小鸟飞的更远——目标
整个游戏过程中需要躲避各种水管——环境
躲避水管的方法是让小鸟用力飞一下——行动
飞的越远，就会获得越多的积分——奖励
二、核心算法(深度强化学习) DQN（先学会1.0的再搞2.0的）
DQN其实可以拆分成Qlearning的主框架加上一些功能：
记忆库：重复学习，缩短时间
神经网络计算Q值：代替原有的表格形式，分析Q值
冻结q_target参数：切断相关性，让训练的效果更好，效率更高
from maze_env import Maze
from RL_brain import DeepQNetwork
def run_maze():
step = 0 # 记录走的步数（先存储记忆库200之后每5步学习一次）
for episode in range(300): #
# 初步的观测值
observation = env.reset() #
while True:
env.render() # 更新环境
# RL choose action based on observation
action = RL.choose_action(observation) # 通过观测值选择一个动作
# RL take action and get next observation and reward
observation_, reward, done = env.step(action) # 把动作放在环境中，获得下一步的情况
# 存储记忆的步骤，放在记忆库里(当前的观测值，当前动作，得到的奖励，下一个观测值)
RL.store_transition(observation, action, reward, observation_)
# 当记忆库满足要啊时开始学习
if (step > 200) and (step % 5 == 0):
RL.learn()
# 赋值为下一个观测值
observation = observation_
# 中断 while 循环
if done:
break
step += 1
# end of game
print('game over')
env.destroy()
if __name__ == "__main__":
# maze game
env = Maze()
# 定义DQN神经网络
RL = DeepQNetwork(env.n_actions, env.n_features,
learning_rate=0.01,
reward_decay=0.9,
e_greedy=0.9,
replace_target_iter=200,
memory_size=2000,
# output_graph=True
)
env.after(100, run_maze)
env.mainloop()
RL.plot_cost()
DQN的神经网络是有两个的，一个是记录最新的，一个是记录最近的动作，奖励等。
tensorflow1.x模型如下：
import numpy as np
import pandas as pd
import tensorflow as tf
np.random.seed(1)
tf.set_random_seed(1)
# Deep Q Network off-policy
class DeepQNetwork:
def __init__(
self,
n_actions, # 输出值的个数
n_features, # 接受观测值的特征
learning_rate=0.01, # 学习率
reward_decay=0.9, # 伽马，奖励衰减率
e_greedy=0.9, # 90%选择最大的，10%探索新的随机值
replace_target_iter=300, # 隔多少步更新一次模型参数
memory_size=500, # 记忆库容量的记录数量的表容量大小
batch_size=32, # 神经网络的批次
e_greedy_increment=None, # 不断缩小随机范围，促进学习
output_graph=False, #
):
self.n_actions = n_actions
self.n_features = n_features
self.lr = learning_rate
self.gamma = reward_decay
self.epsilon_max = e_greedy
self.replace_target_iter = replace_target_iter
self.memory_size = memory_size
self.batch_size = batch_size
self.epsilon_increment = e_greedy_increment
self.epsilon = 0 if e_greedy_increment is not None else self.epsilon_max
# 总学习步数
self.learn_step_counter = 0
# 初始化数组的全零数组长度 [s, a, r, s_]
# n_features * 2两个观测值选择的动作类别("上下左右")，+2其中一个记录记录当前选择的动作，另一个是选择动作得到的价值
self.memory = np.zeros((self.memory_size, n_features * 2 + 2))
# consist of [target_net, evaluate_net] 建立两个神经网络
self._build_net()
t_params = tf.get_collection('target_net_params')
e_params = tf.get_collection('eval_net_params')
self.replace_target_op = [tf.assign(t, e) for t, e in zip(t_params, e_params)]
self.sess = tf.Session()
if output_graph: # tensorboard模型文件的输出
# $ tensorboard --logdir=logs
# tf.train.SummaryWriter soon be deprecated, use following
tf.summary.FileWriter("logs/", self.sess.graph)
# tensorflow会话运行所有初始化变量
self.sess.run(tf.global_variables_initializer())
# 记录每一步误差，为了生成cost误差曲线
self.cost_his = []
# 两个神经网络模型算法的搭建
def _build_net(self):
# ------------------ build evaluate_net ------------------
self.s = tf.placeholder(tf.float32, [None, self.n_features], name='s') # input
self.q_target = tf.placeholder(tf.float32, [None, self.n_actions], name='Q_target') # for calculating loss
with tf.variable_scope('eval_net'):
# c_names(collections_names) are the collections to store variables
c_names, n_l1, w_initializer, b_initializer = \
['eval_net_params', tf.GraphKeys.GLOBAL_VARIABLES], 10, \
tf.random_normal_initializer(0., 0.3), tf.constant_initializer(0.1) # config of layers
# first layer. collections is used later when assign to target net
with tf.variable_scope('l1'):
w1 = tf.get_variable('w1', [self.n_features, n_l1], initializer=w_initializer, collections=c_names)
b1 = tf.get_variable('b1', [1, n_l1], initializer=b_initializer, collections=c_names)
l1 = tf.nn.relu(tf.matmul(self.s, w1) + b1)
# second layer. collections is used later when assign to target net
with tf.variable_scope('l2'):
w2 = tf.get_variable('w2', [n_l1, self.n_actions], initializer=w_initializer, collections=c_names)
b2 = tf.get_variable('b2', [1, self.n_actions], initializer=b_initializer, collections=c_names)
self.q_eval = tf.matmul(l1, w2) + b2
with tf.variable_scope('loss'):
self.loss = tf.reduce_mean(tf.squared_difference(self.q_target, self.q_eval))
with tf.variable_scope('train'):
self._train_op = tf.train.RMSPropOptimizer(self.lr).minimize(self.loss)
# ------------------ build target_net ------------------
self.s_ = tf.placeholder(tf.float32, [None, self.n_features], name='s_') # input
with tf.variable_scope('target_net'):
# c_names(collections_names) are the collections to store variables
c_names = ['target_net_params', tf.GraphKeys.GLOBAL_VARIABLES]
# first layer. collections is used later when assign to target net
with tf.variable_scope('l1'):
w1 = tf.get_variable('w1', [self.n_features, n_l1], initializer=w_initializer, collections=c_names)
b1 = tf.get_variable('b1', [1, n_l1], initializer=b_initializer, collections=c_names)
l1 = tf.nn.relu(tf.matmul(self.s_, w1) + b1)
# second layer. collections is used later when assign to target net
with tf.variable_scope('l2'):
w2 = tf.get_variable('w2', [n_l1, self.n_actions], initializer=w_initializer, collections=c_names)
b2 = tf.get_variable('b2', [1, self.n_actions], initializer=b_initializer, collections=c_names)
self.q_next = tf.matmul(l1, w2) + b2
# 存储记忆
def store_transition(self, s, a, r, s_):
if not hasattr(self, 'memory_counter'):
self.memory_counter = 0 #在记忆库中插入第0行
# 生成数据
transition = np.hstack((s, [a, r], s_))
# 如果超出范围就开始覆盖之前的数据
index = self.memory_counter % self.memory_size
# 记录数据
self.memory[index, :] = transition
# 下一行
self.memory_counter += 1
# 选择动作
def choose_action(self, observation):
# to have batch dimension when feed into tf placeholder
# 因为观测值是一维数据，tensorflow处理需要二位，增加一个维度
observation = observation[np.newaxis, :]
# 如果探索的衰减值在0.9之内，就选择最大的观测值
if np.random.uniform() < self.epsilon:
# forward feed the observation and get q value for every actions
actions_value = self.sess.run(self.q_eval, feed_dict={self.s: observation})
action = np.argmax(actions_value)
else:
# 否则更新随机的新值，为了更好的探索新的走法
action = np.random.randint(0, self.n_actions)
return action
# 学习
def learn(self):
# check to replace target parameters，看需不需要换参数
if self.learn_step_counter % self.replace_target_iter == 0:
self.sess.run(self.replace_target_op)
print('\ntarget_params_replaced\n')
# sample batch memory from all memory
if self.memory_counter > self.memory_size:
sample_index = np.random.choice(self.memory_size, size=self.batch_size)
else:
sample_index = np.random.choice(self.memory_counter, size=self.batch_size)
batch_memory = self.memory[sample_index, :]
# 将两个神经网络的值保存
q_next, q_eval = self.sess.run(
[self.q_next, self.q_eval],
feed_dict={
self.s_: batch_memory[:, -self.n_features:], # fixed params 也就是观测值前四个动作情况
self.s: batch_memory[:, :self.n_features], # newest params 后四个观测值动作情况
})
# change q_target w.r.t q_eval's action
q_target = q_eval.copy()
batch_index = np.arange(self.batch_size, dtype=np.int32)
eval_act_index = batch_memory[:, self.n_features].astype(int)
reward = batch_memory[:, self.n_features + 1]
q_target[batch_index, eval_act_index] = reward + self.gamma * np.max(q_next, axis=1)
"""
For example in this batch I have 2 samples and 3 actions:
q_eval =
[[1, 2, 3],
[4, 5, 6]]
q_target = q_eval =
[[1, 2, 3],
[4, 5, 6]]
Then change q_target with the real q_target value w.r.t the q_eval's action.
For example in:
sample 0, I took action 0, and the max q_target value is -1;
sample 1, I took action 2, and the max q_target value is -2:
q_target =
[[-1, 2, 3],
[4, 5, -2]]
So the (q_target - q_eval) becomes:
[[(-1)-(1), 0, 0],
[0, 0, (-2)-(6)]]
We then backpropagate this error w.r.t the corresponding action to network,
leave other action as error=0 cause we didn't choose it.
"""
# train eval network 输出误差
_, self.cost = self.sess.run([self._train_op, self.loss],
feed_dict={self.s: batch_memory[:, :self.n_features],
self.q_target: q_target})
# 记录误差
self.cost_his.append(self.cost)
# increasing epsilon
self.epsilon = self.epsilon + self.epsilon_increment if self.epsilon < self.epsilon_max else self.epsilon_max
self.learn_step_counter += 1
# cost值曲线可视化
def plot_cost(self):
import matplotlib.pyplot as plt
plt.plot(np.arange(len(self.cost_his)), self.cost_his)
plt.ylabel('Cost')
plt.xlabel('training steps')
plt.show()
