|
发表于 2004-7-30 00:00:00
|
显示全部楼层
关于研究棋谱的分布式计算项目,国外曾经做过尝试,翻译http://www.aspenleaf.com/distributed/distrib-recent.html 页面时发现的,原文如下:
---------------------------------------------------------------------------------------------------
(http://neural-chess.netfirms.com/)
The DIstributed Chess Project tried to create better chess-playing artificial neural networks. The project software implemented a genetic algorithm "to train multi-layer-perceptron neural networks on sets of chess positions with known best continuations (e.g. endgame studies, mate in n moves, white to move and win, ...)." The software was available as a screensaver or command-line client. It allowed the user to "view the status of the evolution any time in terms of computation time, current generation, current best fitness and population diversity" and to "modify the key parameters of the algorithm before a new task starts (e.g. number of generations, population size, number of hidden layers, number of nodes per hidden layer, ...)" if the user wanted to take an active role in shaping its chess-playing neural networks. Volunteers contributed 16 years 57 days of computing time to the project.
-------------------------------------------------------------------------------------------------
大意是:
The DIstributed Chess Project(分布式棋谱研究项目)设法创造出更好的人工神经网络棋手。项目客户端使用了一个基因算法“测试多层感知神经网络学会在一副棋的任意位置上已知的最佳的走法(比如末局研究,n步配对,白手先必赢,...)。”客户端有屏幕保护程序版本和命令行版本。它允许用户手动调整神经网络:“在任何时侯根据估算、当前生成、当前最佳走法和布局差异查看进展状况”并且“在一个新建任务开始之前修改算法的关键参数(比如生成基数,布局范围,隐藏层数,每隐藏层的节点数,...)”。志愿者对项目总计贡献了16年57天的计算时间。
[ Last edited by 碧城仙 on 2005-1-11 at 10:19 PM ] |
|