python强化学习代码示例

以下是一个简单的强化学习代码示例，使用Python和Keras实现了一个基本的Q-learning算法：

import random
import numpy as np
from keras.models import Sequential
from keras.layers import Dense

# 定义游戏环境
num_states = 10
num_actions = 4

# 定义Q-learning算法参数
epsilon = 1.0
epsilon_min = 0.01
epsilon_decay = 0.995
gamma = 0.95
learning_rate = 0.001
batch_size = 32
memory = []

# 定义神经网络模型
model = Sequential()
model.add(Dense(24, input_dim=num_states, activation='relu'))
model.add(Dense(24, activation='relu'))
model.add(Dense(num_actions, activation='linear'))
model.compile(loss='mse', optimizer='adam')

# 定义选择动作的函数
def choose_action(state):
    if np.random.rand() < epsilon:
        return random.randrange(num_actions)
    else:
        q_values = model.predict(state)
        return np.argmax(q_values[0])

# 定义更新模型的函数
def update_model():
    if len(memory) < batch_size:
        return
    batch = random.sample(memory, batch_size)
    for state, action, reward, next_state, done in batch:
        target = reward
        if not done:
            target = reward + gamma * np.amax(model.predict(next_state)[0])
        target_f = model.predict(state)
        target_f[0][action] = target
        model.fit(state, target_f, epochs=1, verbose=0)
    if epsilon > epsilon_min:
        epsilon *= epsilon_decay

# 训练模型
for episode in range(100):
    state = np.random.rand(1, num_states)
    for step in range(100):
        action = choose_action(state)
        next_state = np.random.rand(1, num_states)
        reward = np.random.rand()
        done = False
        memory.append((state, action, reward, next_state, done))
        state = next_state
        update_model()
        if done:
            break

这个示例实现了一个简单的Q-learning算法，使用神经网络模型来估计每个状态-动作对的价值，并使用经验回放来更新模型。在训练过程中，每个episode中的状态是随机生成的，动作是根据当前状态和epsilon-greedy策略选择的，奖励是随机生成的。这个示例只是一个基本的框架，可以根据具体的应用场景进行修改和扩展。