Basic research on brain-inspired information processing systems with integrated computation for recognition and decision making

识别与决策集成计算的类脑信息处理系统基础研究

• 批准号: 24800013
• 负责人: OKU Makito
• 金额: $ 1.25万
• 依托单位: The University of Tokyo
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for Research Activity Start-up
• 财政年份: 2012
• 资助国家: 日本
• 起止时间: 2012-08-31 至 2014-03-31
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-24800013/
• 关键词: 脳型情報処理; 確率推論; ベイズの定理; ダイナミックベイジアンネットワーク; 意思決定; 近似解法; マルコフ連鎖モンテカルロ法; 部分観測マルコフ決定過程

This research aimed for the development of a new computational theory of the brain that integrates two types of brain's functionality, that is, recognition (input-related computation) and decision making (output-related computation). We also tried to establish a fundamental mathematical model based on the theory that may contribute to the realization of highly intelligent information processing systems like the brain.Specifically, we have modeled the computation associated with recognition and decision making as probabilistic inference on a dynamic Bayesian network model in a unified manner. We have also derived the exact solution of the probabilistic inference problem and proposed an approximation method for it.

这项研究旨在开发一种整合两种大脑功能的大脑的计算理论，即识别（与输入相关的计算）和决策（与输出相关的计算）。我们还试图基于理论建立一个基本的数学模型，该理论可能有助于实现高度智能的信息处理系统（例如大脑）。特别是，我们以统一的方式对识别和决策作为对动态贝叶斯网络模型的概率推论的计算进行了建模。我们还得出了概率推理问题的精确解决方案，并为其提出了一种近似方法。

项目成果

Approximated Probabilistic Inference on a Dynamic Bayesian Network Using a Multistate Neural Network

使用多状态神经网络对动态贝叶斯网络进行近似概率推理

• 发表时间: 2013
• 作者: 張建偉;河合由起子;熊本忠彦;白石優旗;張建偉;Makito Oku
• 通讯作者: Makito Oku