Mathematical Study of The Dynamical cell Assembly Hypothesis as the Encoding mechanism of the brain

首席研究员(H.Fujii)和他的合作者提出了一个关于大脑编码机制的新假说--动态细胞组装假说[1]。该假说旨在通过例如神经成像、联合PSTH等，以统一的方式理解显示在大脑中所经历的动力学的最新实验数据，这些实验数据有时难以在传统的基于速率代码假说的网络“理论”的框架内解释，并且还旨在解决关于大脑编码的理论困难，例如绑定问题。这项研究是从实验和理论两个角度证实这一理论的第一步。作为“动态细胞组装假说”的继续，我首先总结了我们对大脑可能实际使用的绑定机制[2]的观点[2]--“关联绑定”，其次，我构建了一个具有两个属性的物体(即两个不同长度和颜色的条)“绑定”的构造性、概念性的计算机模型。该模型由抽象的符合检测神经元构成。该模型的行为是：(I)它表示刺激的完全位置不变性。(Ii)它包含一个在远距离区域处理的属性绑定的动态机制。以上下文相关的方式执行单元组件的动态链接和重新组织。(Iii)多个相关动态正在运行，每个相关动态表示视觉对象之一的统一图像。电池组件的成员通过交叉相关性动态地识别。

The head investigator (H.Fujii) and his collaborators have proposed a new hypothesis on the encoding mechanism of the brain-the dynamical cell assembly hypothesis [1]. The hypothesis is intended to understand in a unified way recent experimental data exhibiting the dynamics undergoing in the brain, by, e.g., neuro-imaging, Joint PSTH and so on, which are sometimes difficult to interpret within the framework of conventional network "theories" based on "rate code hypothesis", and also to resolve theoretical difficulties on the brain coding, e.g., the binding problem. This research is the first step to confirm the theory from both experimental and theoretical standpoints.As a continued work of the "dynamical cell assembly hypothesis", I have, firstly, summarized our standpoint on the binding mechanism which the brain may actually uses [2]-the "correlational scinario of binding", and secondly, I have constructed a constructive, but conceptual computer model of "binding" of two objects with two attributes (i.e., two bars with different lengths and colors). The model is constructed with abstract coincidence detector neurons. The behavior of the model is :(I) It presrves a complete position invariancy of the stimuli.(ii) It contains a dynamic mechanism of binding of attributes which are processed in a distant regions. Dynamical linking and re-organization of cell assemblies are performed in the context dependent manner.(iii) Multiple correlational dynamics are running, each of which represents a unified image of one of the visual object. A member of a cell assembly are identified dynamically by cross correlations.