权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

Learning & Adaptation of Soft Computing Techniques Operating in the Nonstationary Environment

学习

基本信息

批准号：
16500129
负责人：
BABA Norio
金额：
$ 0.77万
依托单位：
Osaka Kyoiku University
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2004
资助国家：
日本
起止时间：
2004 至 2005
项目状态：
已结题

来源：
https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-16500129/
关键词：
Soft Computing Learning Automata Hierarchical Structure Nonstationary Environment Prediction of Stock Price Computer Gaming Fusion of Soft Computing Neural Networks 学習オートマン非定常環境学習適応学習オートマトン

项目摘要

Despite the current matured state concerning the theory and applications of the soft computing techniques, there are still several problems to be settled. One of the most important problems might be : "How to cope with the nonstationary environment ?" Recently, the following idea hit me : "Fusion of learning automata and soft computing techniques may contribute a lot in finding a successful way for solving the above problem."In this research project, I was mainly involved in the research to investigate the learning performance of the hierarchical structure learning automata under the unknown nonstationary multiteacher environment in order to check whether this idea is OK. I was also involved to the research which deals with the real problems that contain rather strong nonlinearity.The following are our research results :1.We proposed a new learning algorithm for the hierarchical structure learning automata operating in the nonstationary multiteacher environment. We proved that the proposed algorithm ensures convergence with probability 1 to the optimal path under a certain nonstationary multiteacher environment.2.In order to compare the learning performance of the proposed algorithm in the nonstationary multiteacher environment with the algorithms DGPA and SE_<RI> (two of the fastest algorithms today), we carried out a rather large numbers of the computer simulations. The simulation results we obtained confirm the effectiveness of the proposed algorithm.3.We have applied neural networks and genetic algorithms to the fields of financial engineering & computer gaming. We succeeded in obtaining the following results :(1)Neural networks are quite helpful in order to predict the Golden Cross and the Dead Cross several weeks before they occur.(2)Fusion of neural networks and genetic algorithms is also quite helpful for finding an appropriate rule to make the Environmental Game much more exciting.

尽管软计算技术的理论和应用已经比较成熟，但仍有一些问题需要解决。其中最重要的问题之一可能是：“如何科普非平稳环境？最近，我突然想到：“学习自动机和软计算技术的融合可能有助于找到解决上述问题的成功方法。”“在这个研究项目中，我主要参与了研究分层结构学习自动机在未知的非平稳多教师环境下的学习性能，以检查这个想法是否可行。本人还参与了对含有较强非线性的真实的问题的研究，取得了以下研究成果：1.提出了一种新的非平稳多教师环境下的层次结构学习自动机的学习算法。证明了在一定的非平稳多教师环境下，该算法以概率1收敛到最优路径。2.为了比较该算法在非平稳多教师环境下的学习性能，与DGPA和SE_<RI>（目前最快的两种算法）进行了大量的计算机仿真。仿真结果验证了算法的有效性。3.将神经网络和遗传算法应用于金融工程和计算机游戏领域。我们成功地获得了以下结果：（1）神经网络对于预测黄金十字架和死十字架的发生有很大的帮助。（2）神经网络和遗传算法的融合也有助于找到一个合适的规则，使环境游戏更令人兴奋。