Development of predictive coding networks for spatial navigation

用于空间导航的预测编码网络的发展 综述：哺乳动物导航使用内部模型来预测时空统计规律 环境位置的顺序。预测编码理论将大脑视为贝叶斯解释器， 计算外部刺激和世界内部模型之间的差值。它越来越多地 理解序列空间信息由集成神经元的时间序列来表示 向海马体开火。无一例外，这些合奏模式已经在 成年动物。少数几项研究测量了单个海马神经元的活动 断奶或幼年动物短暂地探索开阔的野外环境。然而，人们相信，空间体验 由高度压缩的神经元集合放电的时间序列在内部表示 以theta序列、重放和预放的形式表示的唤醒和睡眠状态，在 而不是单个神经元。我们的目标是揭示早期生命发展的原则和阶段 以及作为编码未来导航的先验的基于吸引子的压缩时间序列的成熟 经验(即预测编码)，以及年龄和早期空间经验在这些过程和 在航海学习中。为了实现这一目标，我们开发了新的方法来：1.以毫秒为单位的长期记录 分辨率，来自发育中的海马神经元的大集合(同时高达70个) 大鼠自睁眼后第一天开始自由活动和睡眠；2.揭示和分析预测编码 网络属性；3.控制和限制动物的先验空间体验。成功完成这项工作 提案将提供独特的资源，以帮助了解皮质网络的出现和成熟 用于内部生成的表示，并将提供链接和预测模型来研究 神经发育和精神障碍的神经元集合模式，如精神分裂症和 自闭症谱系障碍。

Development of predictive coding networks for spatial navigation Summary: Mammalian navigation uses internal models to predict the spatial-temporal statistical regularity of the sequence of environmental locations. Predictive coding theories view the brain as Bayesian interpreter that computes the difference between the external stimuli and an internal model of the world. It is increasingly understood that sequential spatial information is represented by temporal sequences of ensemble neuronal firing in the hippocampus. Without exception, these ensemble patterns have been investigated exclusively in adult animals. A small handful of studies measured the activity of individual hippocampal neurons as pre- weanling or juvenile animals briefly explored open-field environments. However, spatial experience is believed to be internally represented by highly compressed temporal sequences of neuronal ensemble firing during awake and sleep states in the form of theta sequences, replay, and preplay, which are expressed within populations rather than single neurons. Our goal is to reveal the principles and stages of early-life development and maturation of attractor-based compressed temporal sequences as priors for encoding future navigation experiences (i.e., predictive coding), and the role age and early spatial experience play in these processes and in navigation learning. To achieve this goal, we develop new methods to: 1. Chronically record, at millisecond resolution, from large ensembles of neurons (up to 70 simultaneously) from the hippocampus in developing freely-behaving and sleeping rats from first day after eye opening; 2. Reveal and analyze predictive coding network properties; 3. Control and restrict animals’ prior spatial experience. Successful completion of this proposal will provide unique resources to help understand the emergence and maturation of cortical networks for internally-generated representations and will provide links and predictive models to study perturbations in neuronal ensemble patterns underlying neurodevelopmental and psychiatric disorders like schizophrenia and autism spectrum disorder.