Reinforcement of the Performances of the Soft Computing Techniques & Utilization of the learning Automaton-Challenges Toward Nonstationary Environment

本文的主要研究成果如下：1）提出了一种新的非平稳多教师环境下的HSLA学习算法。我们已经证明，该算法确保收敛到最佳的行动在一般NME下的概率为1。2）我们推广了DGPA算法用于层次结构的学习自动机模型。在相当一般的条件下，我们证明了广义算法是ε-最优的.我们还提出了几个计算机仿真结果表明，广义算法的有效性的移动的机器人的迷宫通过问题。3）我们进行了大量的计算机仿真，以比较我们的算法的效率在非平稳多教师切换环境中的广义DGPA算法和广义SE_算法。<RI>几乎所有的计算机仿真结果都证实了我们算法的有效性。4）我们尝试了两种进化算法，教育部和FEP使共同游戏令人兴奋。MOEA已被用于生成各种类型的熟练玩家。此外，FEP已被引入，以找出每张卡的点的适当组合。5）提出了一种新的股票交易决策支持系统（DSS），利用神经网络改进了传统的技术分析方法。在所提出的系统中，神经网络被用来预测"黄金十字架"（"死十字架"）发生的几个星期。对东证指数和日经225指数的计算机模拟结果证实了本文方法的有效性。

In the followings, we shall briefly touch upon our research results:1) We have proposed a new learning algorithm of the HSLA operating in the nonstationary multiteacher environment. We have shown that the proposed algorithm ensures convergence to the optimal set of actions with probability 1 under the general NME.2) We have generalized the DGPA algorithm to be used in the hierarchical structure learning automata model. We have shown that the generalized algorithm is ε optimal under the rather general condition. We have also presented several computer simulation results concerning the maze passing problem of the mobile robots that suggest the efficacy of the generalized algorithm.3) We have carried out large numbers of computer simulations in order to compare the efficacy of our algorithm in the nonstationary multiteacher switching environment with those of the generalized DGPA algorithm & the generalized SE_<RI> algorithm. Almost all of the computer simulation results confirm the effectiveness of our algorithm.4) We have tried to utilize two kinds of EAs, ie., the MOEA and the FEP for making the COMMONS GAME exciting. The MOEA has been used for generating various types of skilled players. Further the FEP has been introduced to find out appropriate combinations of the point of each card. We have succeeded in finding highly advanced rules compared with that of the original COMMONS GAME.5) We have proposed a new derision support system (DSS) for dealing stocks which improves the traditional technical analysis by using neural networks. In the proposed system, neural networks are utilized in order to predict the "Golden Cross" ('Dead Cross") several weeks it occurs. Computer simulation results concerning the dealings of the TOPIX and Nikkei-225 have confirmed the effectiveness of our approach.