INFORMATION PROCESSING

信息处理

• 批准号: 3409986
• 负责人: IDAN SEGEV
• 金额: $ 15.07万
• 依托单位: HEBREW UNIVERSITY OF JERUSALEM
• 依托单位国家: 美国
• 财政年份: 1988
• 资助国家: 美国
• 起止时间: 1988-04-01 至 1991-03-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/3409986
• 关键词: Malacostraca; artificial intelligence; axon; biophysics; computer simulation; dendrites; electrophysiology; ganglions; mathematical model; model design /development; neural information processing; neuroanatomy; neurochemistry; neurons; neurophysiology; online computer; saltwater environment; synapses

A new simultaneous experimental and theoretical interdisciplinary approach to study information processing in neuronal networks is suggested. It combines a detailed electro-morphological study at the cellular level with mathematical analysis using a new tool for the Simulation of Neuronal Networks (SONN) which we plan to develop. The idea is to consider neurons as a composition of subunits (e.g., axonal branch point), each performs specific information processing functions (Parnas and Segev, 1979). These functional subunits are first characterized experimentally, then analyzed using detailed partial differential equations (P.D.E) models (e.g., Parnas et al., 1986 a, b, c). Investigating these experimentally based models will eventually allow to abstract the subunits as independent processors or State Machines (S.M), which transform input to output and changes their internal states. SONN will include both P.D.E and the S.M models, the latter will serve to build a whole neuronal network and to analyze its information processing aspects in comparison to the experimental network. We believe that this approach will help to establish a theoretical framework for connecting biophysical mechanisms to the function of the neural-network and will aid to refine and redefine models in Artificial Intelligence. The heart ganglion of the lobster is the experimental network we plan to investigate. The initial modeling will be done with SPICE, a general-purpose electrical network analysis program, which we have adapted for the use of neurobiologists (Segev et al., 1985). However, its limitations and the realization that an appropriate modeling tool is crucial for our purposes, prompted us to propose to develop SONN for performing "on line" theoretical experiments. The emphasis is on developing graphical and mathematical tools to visualize the morphology and the electrical behavior of the modeled system using color codes. We believe this approach will lead to a better understanding of the principles that underlie the function of neuronal networks.

一种新的同时进行的实验和理论交叉学科 研究神经网络中信息处理的方法是 建议。它结合了详细的电形态研究，网址为 使用一种新的工具进行细胞水平的数学分析 我们计划进行的神经元网络(SONN)的模拟 发展。这个想法是把神经元看作是 亚单位(例如，轴突分支点)，每个亚单位执行特定的 信息处理功能(帕纳斯和塞格夫，1979)。这些 首先对功能亚基进行实验表征，然后 使用详细的偏微分方程(P.D.E)进行分析 模型(例如，ParNas等人，1986 a，b，c)。调查这些 基于实验的模型最终将允许抽象 子单元作为独立处理器或状态机(S.M)，其 将输入转换为输出，并更改其内部状态。 SUNN将包括P.D.E和S.M两种型号，后者将 用于构建完整的神经元网络并分析其 信息处理方面与试验性的比较 网络。我们相信，这种做法将有助于建立一种 将生物物理机制与 神经网络的功能和将有助于提炼和 重新定义人工智能中的模型。 龙虾的心脏神经节是我们的实验网络 计划进行调查。初始建模将使用SPICE完成， 一个通用的电力网络分析程序，我们用它 已经适应了神经生物学家的使用(Segev等人，1985年)。 然而，它的局限性和认识到适当的 建模工具对于我们的目的至关重要，促使我们提出 为演绎“在线”理论发展SUNN 实验。重点是开发图形化和 用于可视化形态和电学的数学工具 使用颜色代码的建模系统的行为。我们相信 这种方法将使人们更好地理解这些原则。 这是神经网络功能的基础。