Episodic and associative memory from spatial scaffolds in the hippocampus

• 成果类型: Article
• 署名作者: Chandra, Sarthak; Sharma, Sugandha; Chaudhuri, Rishidev; Fiete, Ila
• 署名单位: Massachusetts Institute of Technology (MIT); University of California System; University of California Davis; University of California System; University of California Davis
• 刊物名称: Nature
• ISSN/ISSBN: 0028-2022
• DOI: 10.1038/s41586-024-08392-y
• 发表日期: 2025-02-20
• 关键词: collective computational properties; impaired recognition memory; place cells; grid cells; neural-networks; storage capacity; INFORMATION; SPACE; FIELDS; map

摘要

Hippocampal circuits in the brain enable two distinct cognitive functions: the construction of spatial maps for navigation, and the storage of sequential episodic memories1, 2, 3, 4-5. Although there have been advances in modelling spatial representations in the hippocampus6, 7, 8, 9-10, we lack good models of its role in episodic memory. Here we present a neocortical-entorhinal-hippocampal network model that implements a high-capacity general associative memory, spatial memory and episodic memory. By factoring content storage from the dynamics of generating error-correcting stable states, the circuit (which we call vector hippocampal scaffolded heteroassociative memory (Vector-HaSH)) avoids the memory cliff of prior memory models11,12, and instead exhibits a graceful trade-off between number of stored items and recall detail. A pre-structured internal scaffold based on grid cell states is essential for constructing even non-spatial episodic memory: it enables high-capacity sequence memorization by abstracting the chaining problem into one of learning low-dimensional transitions. Vector-HaSH reproduces several hippocampal experiments on spatial mapping and context-based representations, and provides a circuit model of the 'memory palaces' used by memory athletes13. Thus, this work provides a unified understanding of the spatial mapping and associative and episodic memory roles of the hippocampus.