Embedded Ensemble Encoding

摘要 我们正在为哺乳动物新皮层开发一种新的嵌入式集成编码（embedded-ensemble encoding，简称CSTR）理论， 细胞和网络实验，并推断信息如何在大脑中处理的一般原则。我们 研究者的组合包括理论家/建模者、实验家/建模者和建模者/ 神经信息学家我们的理论是基于观察到皮层锥体神经元产生 突触诱导的树突平台电位，将单个神经元置于激活状态。这使 使神经元接近阈值，并且还降低膜时间常数，使得激活的细胞PNact可以容易地 并迅速跟随突触输入。我们假设这些活化细胞的集合提供了活化的 集成Eact，嵌入在列的整个单元格中。然后有一个第二次嵌入的合奏 基于Eact的细胞之间的同步尖峰。该两次嵌入的系综被表示为Esync，其中 Esync Eact.区域内的同步尖峰编码则为广泛分布的系综提供了基底 跨区域，允许将多模态特征结合到连贯的对象感知中（基于 同步绑定理论）。同步性理论对解释同步性约束和同步性约束都有直接的影响， 理论和贝叶斯预测编码理论。开发的工具将用于促进其他项目， 我们的最终用户：1.为更详细的分析开发进一步简化的模型（Mihalas）; 2.开发模型， 位置细胞理论（Kubie）; 3.在猕猴中开发新的数据分析和刺激方案，用于脑机 接口开发（弗朗西斯）。 我们建议主要在多尺度模型中工作，以进一步发展非线性理论的细节， 具体的预测。在神经科学中，与物理学不同，对大脑测量的详细预测必须是 通过在模拟中实例化理论获得，这允许实验者识别特定的尺度， 这是理论的一个方面，可以通过他们的实验措施。我们的具体目标是：1。制定一套 第5层锥体细胞的单细胞模型，基于可用的实验数据和形态学，并测试 基底和顶端斜树突上输入的输入/输出活动模式。生成模型预测， 在体外或体内实验中用树突状成像进行测试。2.构建网络并测试点火可变性， 编码密度、信息论信号穿透、图论测量。将进行验证 跨多个模型实例化。具体的实验预测将用于未来的模型验证。 3.通过模型分享、研讨会、教程和 课程将开发和共享的工具包括用于模型参数拟合的遗传算法、背景驱动 和激活输入数据套件，以及特定的细胞和网络模型。

Abstract We are developing a novel embedded-ensemble encoding (EEE) theory for mammalian neocortex to unify data from cell and network experiments, and to infer general principles of how information is processed in the brain. Our combination of investigators includes a theorist/modeler, an experimentalist/modeler and a modeler/ neuroinformatician. Our theory is based on the observation that cortical pyramidal neurons produce synaptically-induced dendritic plateau potentials that place an individual neuron into an activated state. This brings that neuron near to threshold, and also reduces membrane time constant, so that the activated cell PNact can readily and rapidly follow synaptic inputs. We hypothesize that ensembles of these activated cells provide the activated ensemble Eact, embedded in the overall cells of the column. There is then a second embedding of an ensemble based on synchronized spiking among the cells of Eact. This twice-embedded ensemble is denoted as Esync, with Esync Eact. Synchronized spike coding within area then provides the substrate for a broad distributed ensemble across areas that would allow the binding of multimodal features into coherent object perception (based on binding-by-synchrony theory). EEE theory has direct implications for interpretation of both binding-by-synchrony theory, and for theories of Bayesian predictive coding. Developed tools will be used to facilitate other projects through our end-users: 1. developing further reduced models for more detailed analysis (Mihalas); 2. develop models for place cell theory (Kubie); 3. develop new data analysis and stimulation protocols in macaque for use in brain-machine interface development (Francis). We propose to work primarily in a multiscale model both to develop further details of EEE theory, and to make specific predictions. In neuroscience, unlike in physics, detailed predictions for measures in the brain must be obtained by instantiating the theory in simulation, which allows the experimentalist to identify a particular scale and aspect of the theory that is accessible through their experimental measures. Our Specific Aims are: 1. Develop a set of single cell models of Layer 5 pyramidal cells based on available experimental data and morphologies, and test input/output activity patterns for inputs on basilar and apical oblique dendrites. Generate model predictions that can be tested in in vitro or in vivo experiments with dendritic imaging. 2. Build networks and test with firing variability, coding density, information-theoretic signal flow-through, graph-theoretic measures. Verification will be performed across multiple model instantiations. Specific experimental predictions will be made for future model validation. 3. Disseminate theory, models and experimental predictions through model sharing, workshops, tutorials, and courses. Tools to be developed and shared include genetic algorithms for model parameter fitting, background-driving and activation-input data-suites, and specific cell and network models.