RIA: A Self-Organizing Neural Network Model of The PrimaryVisual cortex

RIA：初级视觉皮层的自组织神经网络模型

• 批准号: 9309273
• 负责人: Risto Miikkulainen
• 金额: $ 14.12万
• 依托单位: University of Texas at Austin
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 1993
• 资助国家: 美国
• 起止时间: 1993-06-15 至 1996-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9309273&HistoricalAwards=false
• 关键词: RIA; Self; Organizing; Neural; Network

In a pilot study motivated by recent results on the lateral connectivity and dynamic processes in the mammalian neocortex, a new model of coritcal self-organization called LCSOFM (Laterally Connected Self-Organizing Feature Map) was developed. This research focuses on LCSOFM as a computational model of the primary visual cortex. The main hypothesis is that the lateral connectivity in the visual cortex mediates cooperation and competition necessary for the self-organization and low-level perceptual phenomena. The hypothesis will be tested by building a LCSOFM model of the visual cortex and analyzing its behavior through simulated neuroscience experiments. The LCSOFM model permits computational study of the visual cortex at a new, more fundamental level than earlier models, and allows us to test several new hypotheses about its structure and function. Based on the pilot study, there is reason to believe that the model can verify whether (1) self-organization can be based on cooperation and competition among laterally connected neurons; (2) the observed patterns of long-range lateral connections arise from correlations in neural activity; (3) plasticity of the cortex results from reorganization of lateral connections; (4) tilt illusions arise from competition of neural responses in laterally connected retinotopic maps; (5) gestalt principles can be extracted from visual experience and encoded into lateral connections; and (6) cooperation and competition through lateral connection, combine with binding by phase-locking, can be responsible for segmentation of visual input into coherent wholes.//

在最近关于哺乳动物大脑皮层横向连通性和动态过程的一项先导性研究中，开发了一个新的大脑皮层自组织模型，称为横向连接自组织特征映射(LCSOFM)。本研究以初级视皮层的计算模型--LCSOFM为研究对象。主要的假设是，视皮层的侧向连接调节了自组织和低水平知觉现象所必需的合作和竞争。这一假说将通过建立视觉皮质的LCSOFM模型并通过模拟神经科学实验分析其行为来验证。与以前的模型相比，LCSOFM模型允许在一个新的、更基本的水平上对视觉皮质进行计算研究，并允许我们测试关于其结构和功能的几个新假设。在前人研究的基础上，有理由相信该模型可以验证：(1)侧向连接神经元之间的合作和竞争是否可以建立在自组织的基础上；(2)观察到的长距离侧向连接的模式源于神经活动的相关性；(3)皮质的可塑性是侧向连接重组的结果；(4)斜视错觉是由侧向连接的视网膜图中神经反应的竞争引起的；(5)格式塔原理可以从视觉经验中提取并编码到侧向连接中；(6)通过侧向连接的合作和竞争，以及通过锁相结合的方式，可以负责将视觉输入分割成连贯的整体。