The Geometry of Abstraction in the Hippocampus and Prefrontal Cortex.
The Geometry of Abstraction in the Hippocampus and Prefrontal Cortex.
复制标题
DOI：
10.1016/j.cell.2020.09.031
复制
发表时间：
2020-11-12
期刊：
影响因子：
64.5
通讯作者：
Salzman CD
中科院分区：
文献类型：
--
作者：
Bernardi S;Benna MK;Rigotti M;Munuera J;Fusi S;Salzman CD
维度的诅咒困扰着强化学习和决策模型。抽象过程通过构建描述不同实例共享的特征的变量，降低维数并在新情况下实现泛化来解决这个问题。在这里，我们描述了猴子在执行由不同的隐变量和显变量描述的任务时的神经表征。抽象是使用神经解码器在不用于训练的任务条件下的泛化性能在操作上定义的，这需要神经表征的特定几何形状。前额叶皮质、海马体和模拟神经网络中的神经集合同时在反映抽象的几何中表示多个变量，但仍然允许线性分类器解码大量其他变量（高破碎维度）。此外，这种几何形状随着任务事件和性能的变化而变化。这些发现阐明了大脑和人工系统如何以抽象格式表示变量，同时保留了高破碎维度所带来的优势。不同类型的认知、情绪和行为灵活性——在新情况下的泛化和对复杂输入模式产生多种不同反应的能力——对神经表征提出了不同的要求。这篇论文展示了神经表征的几何如何对阐明大脑如何支持这些形式的灵活行为至关重要。
The curse of dimensionality plagues models of reinforcement learning and decision-making. The process of abstraction solves this by constructing variables describing features shared by different instances, reducing dimensionality and enabling generalization in novel situations. Here we characterized neural representations in monkeys performing a task described by different hidden and explicit variables. Abstraction was defined operationally using the generalization performance of neural decoders across task conditions not used for training, which requires a particular geometry of neural representations. Neural ensembles in prefrontal cortex, hippocampus, and simulated neural networks simultaneously represented multiple variables in a geometry reflecting abstraction but that still allowed a linear classifier to decode a large number of other variables (high shattering dimensionality). Furthermore, this geometry changed in relation to task events and performance. These findings elucidate how the brain and artificial systems represent variables in an abstract format while preserving the advantages conferred by high shattering dimensionality. Different types of cognitive, emotional and behavioral flexibility – generalization in novel situations and the ability to generate many different responses to complex patterns of inputs -place different demands on neural representations. This paper shows how the geometry of neural representations can be critical for elucidating how the brain supports these forms of flexible behavior.
登录
查看更多内容
影响因子：
16.2
作者：
Kumaran, Dharshan;Summerfield, Jennifer J.;Hassabis, Demis;Maguire, Eleanor A.
通讯作者：
Maguire, Eleanor A.
影响因子：
5
作者：
Dietterich, TG
通讯作者：
Dietterich, TG
影响因子：
2.5
作者：
Isik, Leyla;Tacchetti, Andrea;Poggio, Tomaso
通讯作者：
Poggio, Tomaso
DOI：
10.1523/jneurosci.2753-12.2013
发表时间：
2013-02-27
期刊：
The Journal of neuroscience : the official journal of the Society for Neuroscience
影响因子：
--
作者：
Barak O;Rigotti M;Fusi S
通讯作者：
Fusi S
影响因子：
19.9
作者：
King JR;Dehaene S
通讯作者：
Dehaene S