Coordination of Self-Organization and External Intelligence

自组织与外部智能的协调

• 批准号: 09680379
• 负责人: MATSUYAMA Yasuo
• 金额: $ 2.05万
• 依托单位: Waseda University
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Scientific Research (C)
• 财政年份: 1997
• 资助国家: 日本
• 起止时间: 1997 至 1998
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/grant/KAKENHI-PROJECT-09680379/
• 关键词: Self-organization; learning algorithm; feature map; morphing; external intelligence; virtual reality; 多重降下競合学習

In self-organizing systems, a massively large number of information processing elements with modifiable states work together in coordination. Such a total system can reveal complex and sophisticated information processing. This research project utilized this ability in joint data compression and virtual movie generation. The following methods and results were obtained by this study.(1) Multiple descent cost competition is applied to obtain a new learning and self-organization algorithm. This algorithm generates a weight vector feature map and a group vector feature map simultaneously.(2) Digital still image is data-compressed by the multiple descent cost competition. This created a set of deformable color region patterns and a group mesh pattern.(3) The grouping pattern was deformed according to the specification of external intelligence so that virtual digital movie can be generated.(4) Depth information was further added so that virtual 3D movie can be generated.The above new methods and results can be understood to improve computer human interface.

在自组织系统中，大量具有可修改状态的信息处理元件协同工作。这样一个完整的系统可以揭示复杂和复杂的信息处理。本研究计划利用此能力，在联合数据压缩和虚拟电影生成。通过本研究获得了以下方法和结果。(1)采用多下降成本竞争的方法，得到一种新的学习自组织算法。该算法同时生成权向量特征图和组向量特征图。(2)数字静止图像的数据压缩是通过多重下降成本竞争来实现的。这创建了一组可变形颜色区域图案和一组网格图案。(3)根据外部智能的规范对分组模式进行变形，从而生成虚拟数字电影。(4)进一步加入深度信息，生成虚拟三维动画，这些新方法和结果对改善人机界面具有一定的借鉴意义。

项目成果

Yasuo MATSUYAMA: "The α-EM algorithm" Lecture Notes in Computer Science. 1240. 483-492 (1997)

Yasuo MATSUYAMA：“α-EM 算法”计算机科学讲义。1240. 483-492 (1997)

Yasuo MATSUYAMA et al.: "Prior probability weights and neural network learning" Proc.Int.Conf.Neural Info.Processing. 1. 267-270 (1997)

Yasuo MATSUYAMA 等人：“先验概率权重和神经网络学习”Proc.Int.Conf.Neural Info.Processing。

Y.Matsuyama: "Non-logarithmic information measures, α-EM algorithms and speedup of learning" Proc.IEEE Int.Symp.Information Theory. 1. 385 (1998)

Y.Matsuyama：“非对数信息测量、α-EM 算法和学习加速”Proc.IEEE Int.Symp.信息理论。1. 385 (1998)

Y.Matsuyama,et al.: "A hierarchy from α-EM algorithm to vector quantization and self-organization" Proc.Int.Conf.Neural Info.Processing. 1. 233-238 (1998)

Y. Matsuyama 等人：“从 α-EM 算法到矢量量化和自组织的层次结构”Proc.Int.Conf.Neural Info.Processing。1. 233-238 (1998)

Yasuo MATSUYAMA: "WEM algorithms and probabilistic learning" Proc.ISCIE Stochastic System Sympo.1. 1-12 (1997)

Yasuo MATSUYAMA：“WEM 算法和概率学习”Proc.ISCIE Stochastic System Sympo.1。