Analysis of Perceptual Alternation Phenomena and Artificial Consciousness Model based on Non-standard Computation

基于非标计算的感知交替现象与人工意识模型分析

基本信息

批准号：
13680458
负责人：
MATSUI Nobuyuki
金额：
$ 2.3万
依托单位：
HIMEJI INSTITUTE OF TECHNOLOGY (University of Hyogo from 2004.4)
依托单位国家：
日本
项目类别：
Grant-in-Aid for Scientific Research (C)
财政年份：
2001
资助国家：
日本
起止时间：
2001 至 2003
项目状态：
已结题

项目摘要

In this research, we dealt with the perceptual-alternation phenomena and a theoretical scheme for Artificial Consciousness based on Non-standard Computation. In order to construct a new framework for this perceptual-alternation, we have developed new computation methods, describing the cohesiveness of distribution and synthesis inherent in a neural network. This is called Non-standard computation models : Qubit neural network, Quaternion neural network, Fault-tolerant Asynchronous Cellular Automata, Neural Chaos model, Coupled Oscillatory neural network, Neutral variation Genetic Algorithm, Coupled Gradient Descent method, Stochastic Resonance neural network, Block-Matching based Self-Organizing Map and so on. Discussing them from a theoretical and unified point of view, we investigated their information processing abilities by processing concrete problems as bench mark tests : n-bits parity check problems, general function identification problems, image compression problems, non-holonomic control problems, etc. As a result, we clarified that our non-standard computation models have better performances than conventional models in processing various problems. We have not only established the framework of non-standard computing methods, but also constructed the non-standard computation description on the binding problem. We, furthermore, indicated the existence of quantal state in brain by means of our analysis of the perceptual alternation as a change of the mode of consciousness. This will contribute to the construction of models of computational Artificial Consciousness for future works.

在这项研究中，我们研究了感知交替现象和基于非标准计算的人工意识的理论方案。为了构建这种感知交替的新框架，我们开发了新的计算方法，描述神经网络固有的分布和合成的内聚性。这称为非标准计算模型：量子位神经网络、四元数神经网络、容错异步元胞自动机、神经混沌模型、耦合振荡神经网络、中性变异遗传算法、耦合梯度下降法、随机共振神经网络、块-基于匹配的自组织映射等。我们从理论和统一的角度对它们进行讨论，通过处理具体问题作为基准测试来考察它们的信息处理能力：n位奇偶校验问题、通用函数识别问题、图像压缩问题、非完整控制问题等。结果，我们澄清了我们的非标准计算模型在处理各种问题时比传统模型具有更好的性能。我们不仅建立了非标准计算方法的框架，而且构建了绑定问题的非标准计算描述。此外，我们通过对知觉交替作为意识模式变化的分析，表明了大脑中量子态的存在。这将有助于为未来的工作构建计算人工意识模型。