REMERGE: A complementary learning systems account of the hippocampal contribution to generalization

Kumaran, D. 1 & McClelland, J. 2

1 Universtiy College London
2 Stanford University

We present a perspective on the role of the hippocampal system in generalization, instantiated in a computational model called REMERGE (Recurrency, and Episodic Memory Results in Generalization). We expose a fundamental, but neglected, tension between prevailing computational perspectives of hippocampal function and empirical and theoretical support for its role in generalization and flexible relational memory. Our account provides a means by which to resolve this conflict, by demonstrating that the basic representational scheme proposed by Complementary Learning Systems theory, relying on orthogonalized codes in the hippocampus, can support efficient generalization, as long as there is recurrence rather than unidirectional flow within the hippocampal circuit, or between the hippocampus and neocortex. We suggest that recurrence expands the generalization capacities of classical exemplar-based models of memory and allows the hippocampus to support efficient generalization through recurrent similarity computation, a process which involves the interaction of related episodic experiences within a dynamically created memory space.