Creative information processing systems based on non-standard computation toward brain potentiality

在本研究中，我们试图寻找基于非标准计算的创造性思维方法，以感知交替现象等大脑潜能为基础，并通过对这些方法的系统应用来研究这种创造性效应的产生。在我们对创造性信息处理系统的研究中，我们获得了以下主要成果：1。在离散状态和离散时间用四元数描述的Hopfield神经网络中，我们发现了多重特征的关联。我们证实了量子比特神经网络在彩色夜视和数据挖掘应用中的有效性。我们开发了具有高度非平稳环境适应性的块学习SOM系统。我们利用多层双向联想记忆消除了多稳定感知模型的有效性。我们从一个强化学习的方案中推导出智能体的群集行为，并阐明了在迷宫世界问题中使用混沌探索的强化学习的效果。我们在异步元胞自动机中构造了缺陷容忍度算法，并证明了缺陷细胞是通过在细胞空间中移动的随机苍蝇的配置来检测和隔离的。结果表明，我们的创造性信息处理系统在处理各种问题时比传统的系统有更好的性能。我们已经展示了许多卓有成效的成果，因此我们认为我们的研究目标已经实现。但仍有许多课题有待探索，如Clifford代数方法、量子比特神经元状态的Bloch表示及其在细胞神经网络中的应用等。这些是我们未来的作品。

In this research, we tried to find the methods of creative thinking based on non-standard computation toward brain potentiality such as perceptual alternation phenomena and investigated the emergence of this creative effect through system applications to these methods. In our study of the creative information processing systems, we have obtained the following main results:1. We found the multiplet as the characteristics of association in discrete states and discrete time Hopfield neural network described by quaternion.2. We confirmed the efficiency of the qubit neural network in the color night vision and data-mining applications.3. We developed the block learning SOM system with high adaptability under non-stationary environments.4. We cleared the efficiency of the multi-stable perception model by multi-layered bi-directional associative memory.5. We derived flocking behaviors of agents from a scheme of reinforcement learning, and also clarified the effect of reinforcement learning using chaotic exploration in the maze world problem.6. We constructed the algorithms on defect-tolerance in the asynchronous cellular automaton and showed that defective cells are detected and isolated by configurations of random flies that move around in cellular space.As a result, we clarified that our creative information processing systems have better performances than conventional ones in processing various problems. We have presented a lot of our fruitful results so that we consider our research aims have been achieved. However, many subjects remain to be explored, for examples, Clifford algebra approach, Bloch representation on qubit neuron state and its application to cellular neural network, etc. These are our future works.