基于非负稀疏编码的位置细胞反馈环路学习模型  

A Place Cell Recurrent Loop Learning Model Based on Non-negative Sparse Coding

作　　者：任梦辉 王东署 REN Menghui;WANG Dongshu(School of Electrical and Information Engineering,Zhengzhou University,Zhengzhou 450001,China;Innovation Center of Intelligent Systems,Longmen Laboratory,Luoyang 471000,China)

机构地区：[1]郑州大学电气与信息工程学院,河南郑州450001 [2]龙门实验室智能系统科创中心,河南洛阳471000

出　　处：《郑州大学学报（理学版）》2025年第1期31-39,共9页Journal of Zhengzhou University:Natural Science Edition

基　　金：国家自然科学基金项目(62173309);河南省重点研发专项(231111211600)。

摘　　要：为了探究大脑导航编码的神经机制,聚焦内嗅皮层与海马体之间的神经连接进行模型研究。生理学证据显示,内嗅皮层与海马体之间存在显著的反馈回路连接,两者的空间编码细胞在导航行为中表现出高度关联性。基于这一基础,建立了反馈循环网络模型,将内嗅皮层的栅格细胞与弱空间细胞作为网络输入,连接到海马体的位置细胞与颗粒细胞,并采用非负稀疏编码进行学习。实验结果表明:该反馈学习模型可以快速捕获细胞的空间调谐特性,仅使用弱空间细胞作为输入,也可以通过反馈环路学习到海马位置细胞对空间的单峰选择性,说明反馈编码机制在优化空间表示中发挥着关键作用。总之,该模型可能是大脑导航系统生成精确空间编码的重要细胞机制之一。To explore the neural mechanisms of brain navigation encoding,focus was placed on the neural connections between the entorhinal cortex and the hippocampus for model research.Physiological evidence showed that significant feedback loop connections existed between the entorhinal cortex and the hippocampus,with the spatial encoding cells of both being highly correlated in navigational behavior.Based on this foundation,a feedback loop network model was established,where grid cells and weak spatial cells from the entorhinal cortex were taken as network inputs,connected to place cells and granule cells in the hippocampus,and non-negative sparse coding was employed for learning.Experimental results indicated that the feedback learning model could rapidly capture the spatial tuning properties of these cells.Even when only weak spatial cells were used as inputs,the hippocampal place cells′unimodal selectivity to space could be learned through the feedback loop,suggesting that the feedback encoding mechanism played a key role in optimizing spatial representation.In summary,the model might be one of the important cellular mechanisms for generating precise spatial encoding in the brain′s navigation system.

关 键 词：内嗅皮层 海马体 栅格细胞 位置细胞 反馈循环 非负稀疏编码 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]