基于自博弈深度强化学习的黑白棋系统.zip资源-CSDN文库

共52个文件

pyc：26个

py：22个

md：2个

版权申诉

166 浏览量 2024-05-09 09:53:18 上传评论收藏 57.79MB ZIP 举报

基于自博弈深度强化学习的黑白棋系统深度学习（Deep Learning，简称DL）是机器学习（Machine Learning，简称ML）领域中一个新的研究方向，其目标是让机器能够像人一样具有分析学习能力，识别文字、图像和声音等数据。深度学习通过学习样本数据的内在规律和表示层次，使机器能够模仿视听和思考等人类活动，从而解决复杂的模式识别难题。深度学习的核心是神经网络，它由若干个层次构成，每个层次包含若干个神经元。神经元接收上一层次神经元的输出作为输入，通过加权和转换后输出到下一层次神经元，最终生成模型的输出结果。神经网络之间的权值和偏置是神经网络的参数，决定了输入值和输出值之间的关系。深度学习的训练过程通常涉及反向传播算法，该算法用于优化网络参数，使神经网络能够更好地适应数据。训练数据被输入到神经网络中，通过前向传播算法将数据从输入层传递到输出层，然后计算网络输出结果与实际标签之间的差异，即损失函数。通过反向传播算法，网络参数会被调整以减小损失函数值，直到误差达到一定的阈值为止。深度学习中还包含两种主要的神经网络类型：卷积神经网络（Convolutional Neural Networks，简称CNN）和循环神经网络（Recurrent Neural Networks，简称RNN）。卷积神经网络特别擅长处理图像数据，通过逐层卷积和池化操作，逐步提取图像中的高级特征。循环神经网络则适用于处理序列数据，如文本或时间序列数据，通过捕捉序列中的依赖关系来生成模型输出。深度学习在许多领域都取得了显著的成果，包括计算机视觉及图像识别、自然语言处理、语音识别及生成、推荐系统、游戏开发、医学影像识别、金融风控、智能制造、购物领域、基因组学等。随着技术的不断发展，深度学习将在更多领域展现出其潜力。在未来，深度学习可能会面临一些研究热点和挑战，如自监督学习、小样本学习、联邦学习、自动机器学习、多模态学习、自适应学习、量子机器学习等。这些研究方向将推动深度学习技术的进一步发展和应用。

资源推荐

资源详情

资源评论

收起资源包目录

基于自博弈深度强化学习的黑白棋系统.zip （52个子文件）

content

doc

readme.md 2KB

unispac_21.pdf 855KB

reversi.py 8KB

engines

human.py 2KB

__init__.py 1KB

noorder.py 14KB

greedy.py 1KB

random.pyc 881B

alpha

chessGame

reversiGame.py 4KB

chessBoard.py 6KB

__pycache__

reversiGame.cpython-36.pyc 3KB

chessBoard.cpython-36.pyc 5KB

NetWork

networkFrame.py 4KB

gameNetwork.py 3KB

__pycache__

gameNetwork.cpython-36.pyc 2KB

networkFrame.cpython-36.pyc 4KB

tranningFrame

competition.py 1KB

selfCompetition.py 7KB

__pycache__

competition.cpython-36.pyc 1KB

selfCompetition.cpython-36.pyc 5KB

MCTS

mctAgent.py 1KB

MCTS.py 4KB

__pycache__

MCTS.cpython-36.pyc 3KB

mctAgent.cpython-36.pyc 1KB

model

best.pth.tar 61.22MB

myUtil

util.py 81B

__pycache__

util.cpython-36.pyc 429B

eona.pyc 14KB

__init__.pyc 2KB

simple.py 12KB

simple2.py 13KB

order.py 14KB

random.py 461B

eona.py 16KB

__pycache__

human.cpython-36.pyc 2KB

eona.cpython-36.pyc 12KB

greedy.cpython-36.pyc 1KB

unispac_21.cpython-36.pyc 2KB

__init__.cpython-35.pyc 1KB

agent.cpython-36.pyc 2KB

order.cpython-36.pyc 12KB

noorder.cpython-36.pyc 11KB

random.cpython-35.pyc 766B

__init__.cpython-36.pyc 1KB

human.cpython-35.pyc 2KB

nonull.cpython-36.pyc 11KB

greedy.cpython-35.pyc 1KB

random.cpython-36.pyc 717B

nonull.py 14KB

unispac_21.py 1KB

readme.md 2KB

board.py 7KB

# How to run? Dependencies : numpy, pytorch1.0.1.post2 (The newest edition download from <https://pytorch.org/> ) cuda10.0 CUDA 10.0.130 CUDNN 7.5.0.56 In the root directory, there are two source files : board.py and reversi.py, which are offered by the origin engine frame. And a single directory : engines. You can get my engine interface "unispac_21.py" in the "engines" directory. Run the engine by input : python3 reversi.py -a eona -b unispac_21 Then you can see the match play between my engine and the example engine. Note that in oder to take best use of the time resource, each step will cost about 30 seconds.. If you want to see the result more quickly, open the source file : "unispac_21.py" and modify the args.simCntOfMCT. 100~500 is recommended.. !!! Last but not least.. If you want to transplant my engine to another game environment. Don't forget to take along the directory "alpha" under "engines" togther. The directory "alpha" offers the key module of my engine.. And as there may be a problem of path.. It is recommended to organize the game just like the example frame.. Put all the engine files in the engiles directory. Put my "alpha" directory under the engiens dir.. And the reversi.py is under the root directory.. # About author Developer : Qi Xiangyu(漆翔宇) Sid : 3170104557 GroupNumber : 21 Major : Computer Science and Technology Tel : 17342017090 Email : unispac@zju.edu.cn QQ : 416162623 Please contact with me if you have encountered any problems when running it.. The report will be handed in before 8:00 am on 4.21 as required..

评论收藏

内容反馈

版权申诉