Episodic and associative memory from spatial scaffolds in the hippocampus

Hippocampal circuits in the brain enable two distinct cognitive functions: the construction of spatial maps for navigation and the storage of sequential episodic memories. This dual role remains an enduring enigma. While there have been advances in modeling spatial representations in the hippocampus, 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 factorizing content storage from the dynamics of generating error-correcting stable states. Unlike existing neural memory models, which exhibit a memory cliff, the circuit (which we call Vector-HaSH, Vector Hippocampal Scaffolded Heteroassociative Memory) exhibits a graceful tradeoff between number of stored items and detail. Next, we show that structured internal scaffold states are essential for constructing episodic memory: they enable high-capacity sequence memorization by abstracting the chaining problem into one of learning low-dimensional transitions. Finally, we show that previously learned spatial sequences in the form of cortico-hippocampal location-landmark associations can in turn be used as larger scaffolds and associated with neocortical inputs for a high-fidelity one-shot memory, providing the first circuit model of the "memory palaces" used in the striking feats of memory athletes.