Posani Lorenzo, Wang Shuqi, Muscinelli Samuel P, Paninski Liam, Fusi Stefano
Zuckerman Institute, Columbia University, New York, NY, USA.
School of Computer and Communication Sciences, EPFL, Street, Lausanne, Switzerland.
bioRxiv. 2025 Feb 13:2024.11.15.623878. doi: 10.1101/2024.11.15.623878.
A long-standing debate in neuroscience concerns whether individual neurons are organized into functionally distinct populations that encode information differently ("categorical" representations [1-3]) and the implications for neural computation. Here, we systematically analyzed how cortical neurons encode cognitive, sensory, and movement variables across 43 cortical regions during a complex task (14,000+ units from the International Brain Laboratory public Brain-wide Map data set [4]) and studied how these properties change across the sensory-cognitive cortical hierarchy [5]. We found that the structure of the neural code was scale-dependent: on a whole-cortex scale, neural selectivity was categorical and organized across regions in a way that reflected their anatomical connectivity. However, within individual regions, categorical representations were rare and limited to primary sensory areas. Remarkably, the degree of categorical clustering of neural selectivity was inversely correlated to the dimensionality of neural representations, suggesting a link between single-neuron selectivity and computational properties of population codes that we explained in a mathematical model. Finally, we found that the fraction of linearly separable combinations of experimental conditions ("Shattering Dimensionality" [6]) was near maximal across all areas, indicating a robust and uniform ability for flexible information encoding throughout the cortex. In conclusion, our results provide systematic evidence for a non-categorical, high-dimensional neural code in all but the lower levels of the cortical hierarchy.
神经科学领域长期存在的一个争论是，单个神经元是否被组织成功能上不同的群体，这些群体以不同的方式编码信息（“分类”表征[1-3]）以及这对神经计算的影响。在这里，我们系统地分析了在一项复杂任务中，43个皮质区域的皮质神经元如何编码认知、感觉和运动变量（来自国际大脑实验室公开的全脑图谱数据集的14000多个单元[4]），并研究了这些特性如何在感觉-认知皮质层级结构中变化[5]。我们发现神经编码的结构是尺度依赖的：在整个皮质尺度上，神经选择性是分类性的，并且以反映其解剖学连接性的方式在各区域间组织起来。然而，在单个区域内，分类表征很少见，且仅限于初级感觉区域。值得注意的是，神经选择性的分类聚类程度与神经表征的维度呈负相关，这表明了单个神经元选择性与群体编码的计算特性之间的联系，我们在一个数学模型中对此进行了解释。最后，我们发现实验条件的线性可分组合的比例（“打破维度”[6]）在所有区域都接近最大值，这表明整个皮质在灵活信息编码方面具有强大且一致的能力。总之，我们的结果为除皮质层级结构较低水平之外的所有区域中存在非分类的、高维神经编码提供了系统证据。
