君子不器

import torchimport torch.nn as nnimport torch.nn.functional as Fimport numpy as npimport gymimport time##################### hyper parameters ####################EPISODES = 200EP_STEPS = 200LR_ACTOR = 0.001LR_CRITIC = 0.002GAMMA = 0.9TAU = 0

# -- coding: utf-8 --# 单用户import torchimport torch.nn as nnimport torch.nn.functional as Fimport numpy as npimport gym# 超参数BATCH_SIZE = 32LR = 0.01 # learning rateEPSILON = 0.9 # 最优选择动作百分比GAMMA = 0.9

"""Deep Deterministic Policy Gradient (DDPG), Reinforcement Learning.DDPG is Actor Critic based algorithm.Pendulum example.View more on my tutorial page: https://morvanzhou.github.io/tutorials/Using:tensorflow 1.0gym 0.8.0"""######################

import gymimport tensorflow as tfimport numpy as npimport randomfrom collections import deque# Hyper Parameters for DQNGAMMA = 0.9 # discount factor for target QINITIAL_EPSILON = 0.5 # starting value of epsilonFINAL_EPSILON = 0.01 # final value ..

import gymimport tensorflow as tfimport numpy as npimport randomfrom collections import deque# Hyper ParametersGAMMA = 0.95 # discount factorLEARNING_RATE=0.01class Actor():# PI def __init__(self, env, sess): # init some parameters..

######################################################################## Copyright (C) ## 2016 - 2019 Pinard Liu(liujianping-ok@163.com) ## https://www.cnblogs.com/pinard .

第一个文件：my_miniimagenet_train.pyimport osos.environ['CUDA_VISIBLE_DEVICES']='0'import torchfrom my_MiniImagenet import MiniImagenetimport numpy as npfrom my_meta import Metaimport argparsefrom torch.utils.data import DataLoaderdef main(): .

主要的组件：TD，Q_leaning，神经网络表示Q函数，贝尔曼方程。from __future__ import print_functionimport tensorflow as tfimport numpy as npimport cv2import syssys.path.append("game/")import game.wrapped_flappy_bird as gameimport random#游戏名GAME = 'flappy bird'ACTIONS = 2

import osos.environ["CUDA_VISIBLE_DEVICES"] = "1"import tensorflow as tfimport numpy as npimport gymimport matplotlib.pyplot as pltRENDER = False#利用当前策略进行采样，产生数据class Sample(): def __init__(self,env, policy_net): self.env = env

pytorch和运行代码版本不兼容的问题。从 pytorch1.1退到pytorch1.0就可以了。pytorch的安装，可以参考我之前的博客。

sarsa：from yuanyang_env_td import YuanYangEnvimport numpy as npimport randomclass TD_RL: def __init__(self, yuanyang): self.gamma = yuanyang.gamma self.yuanyang = yuanyang # 值函数的初始值 self.qvalue = np.zeros((len(self.y

import pygameimport timeimport randomimport numpy as npimport matplotlib.pyplot as pltfrom yuanyang_env_mc import YuanYangEnvclass MC_RL: def __init__(self, yuanyang): #行为值函数的初始化 self.qvalue = np.zeros((len(yuanyang.states),len(y.

策略评估，策略改进import pygamefrom load import *import mathimport timeimport randomimport numpy as npclass YuanYangEnv: def __init__(self): self.states=[] for i in range(0,100): self.states.append(i) self.actions = [

main.pyimport pygameimport numpy as npfrom load import *import randomclass YuanYangEnv: def __init__(self): self.states = []# 状态 for i in range(0,100): self.states.append(i) self.actions = ['e', 's', 'w', 'n']#

"""在多臂赌博机的实际环境下测试贪心策略，玻尔兹曼策略，UCB策略。第一步：要定义一些多臂赌博机系统的基本信息。第二步：训练 a,选动作，以怎么样的方式选策略。例如：贪心策略，玻尔兹曼策略，UCB策略。 b,将动作发给环境，环境给出即时奖励。第三步：做图，x：玩家玩游戏的总的次数，y：累加奖励。 i, sum_reward(i)"""# 第一步：要定义一些多臂赌博机系统的基本信息。import matplotl...