Computational and Circuit Mechanisms Underlying Rapid Learning

项目摘要/摘要 哺乳动物的大脑具有非凡的存储和检索信息的能力。详细的记忆可以 只需一次曝光就能形成，这些记忆可以保留几十年。这种能力是 内侧颞叶结构，包括海马体受损后受损 以及邻近的皮质。在过去的十年里，许多研究都强调了 海马和新皮质，特别是前额叶皮质(PFC)和后顶叶皮质(PPC)，AS 在巩固记忆方面起着至关重要的作用。然而，支持内存的电路机制 巩固还没有被很好地理解，特别是在灵长类动物的大脑中。记忆力受损是一个重要的 阿尔茨海默病、颞叶癫痫、抑郁症和精神分裂症等疾病的组成部分 这些措施总共影响了2000多万美国人。我们的长远目标是为更好的 了解记忆过程背后的神经机制，以便让我们更接近 为这些残疾患者开发新的治疗方法。心理学理论和行为研究 认为信息的快速单次积累是由先验知识促进的，这是一种认知地图 或者“心理图式”，它提供了一个框架，可以在上面吸收新的信息。这个概念是 与理解记忆巩固服务中潜在的海马区-新皮质相互作用有关。 这里提出的实验将直接检查海马体、PFC和PPC中的神经回路 支持图式开发和新的学习。这一U-19计划的总体目标是开发一种 支持快速学习的电路机制的全面理论。为了实现这些目标，我们将 利用多个实验室的研究框架，进行雄心勃勃的努力，需要以下多个领域 专业知识，以我们的团队成员为例。我们的团队工作围绕四个研究项目进行组织，每个项目 由数据科学和管理核心支持。通过猴子和人类的平行项目，我们将 在海马区、PFC区和PPC区同时进行大规模记录，以评估调制 在图式开发和新的联想和分类学习期间的跨区域连通性。 互补的理论方法将整合人类和非人类的大规模电路建模 基于测量的介观连通性和训练递归神经网络的灵长类大脑执行 认知任务。我们将测试在模式实例化过程中对任务结构进行编码的假设 在连接权重空间中以低维结构的形式，这体现在低- 神经动力学的维度子空间。在新的学习过程中，系统受益于图式的缩小 权重参数搜索，从而加快学习速度。我们假设这一过程在 区域间动态相互作用的水平。综上所述，本计划下提出的实验将 对快速学习的神经机制进行全面、跨物种的调查。

PROJECT SUMMARY/ABSTRACT The mammalian brain has a remarkable ability to store and retrieve information. Detailed memories can be formed after as little as one exposure, and those memories can be retained for decades. This ability is compromised following damage to structures located in the medial temporal lobe, including the hippocampus and the adjacent cortex. Over the past decade, many studies have highlighted interactions between the hippocampus and neocortex, in particular, the prefrontal cortex (PFC) and posterior parietal cortex (PPC), as having an essential role in memory consolidation. However, the circuit mechanisms that support memory consolidation are not well-understood, particularly in the primate brain. Impaired memory is an important component of diseases such as Alzheimer's disease, temporal lobe epilepsy, depression, and schizophrenia that collectively affect over twenty million Americans. Our long-range goal is to contribute to a better understanding of the neural mechanisms that underlie memory processes, in order to bring us closer to developing new therapies for these disabled patients. Psychological theories and behavioral studies have suggested that rapid, single-trial accumulation of information is facilitated by prior knowledge, a cognitive map or “mental schema” that provides a framework onto which new information can be assimilated. This concept is relevant for understanding potential hippocampal-neocortical interactions in the service of memory consolidation. The experiments proposed here will directly examine the neural circuits in the hippocampus, PFC, and PPC that support schema development and new learning. The overall goal of this U-19 Program is to develop a comprehensive theory of the circuit mechanisms that support rapid learning. To achieve these goals, we will make use of a multi-laboratory research framework with an ambitious effort that requires multiple areas of expertise, exemplified by our team members. Our team effort is organized around four Research Projects, each supported by Data Science and Administrative Cores. Through parallel projects in monkeys and humans, we will perform large-scale recordings simultaneously across the hippocampus, PFC and PPC to assess modulations in cross-regional connectivity during schema development and new association and categorization learning. Complementary theoretical approaches will integrate large-scale circuit modeling of the human and nonhuman primate brain based on measured mesoscopic connectivity and training recurrent neural networks to perform cognitive tasks. We will test the hypothesis that in the course of schema instantiation, a task structure is encoded in the form of a low-dimensional structure in the space of connection weights, which is reflected in a low- dimensional subspace of neural dynamics. During new learning, the system benefits from the schema to narrow weight parameter search, thereby speeding up learning. We hypothesize that this process is observable at the level of dynamical inter-areal interactions. Taken together, the experiments proposed under this Program will provide a comprehensive, cross-species investigation of the neural mechanisms of rapid learning.