Correcting the Hebbian Mistake: Toward a Fully Error-Driven Hippocampus

Abstract

The hippocampus plays a critical role in the rapid learning of new episodic memories. Many computational models propose that the hippocampus is an autoassociator that relies on Hebbian learning (i.e., “cells that fire together, wire together”). However, Hebbian learning is computationally suboptimal as it modifies weights unnecessarily beyond what is actually needed to achieve effective retrieval, causing more interference and resulting in a lower learning capacity. Our previous computational models have utilized a powerful, biologically plausible form of error-driven learning in hippocampal CA1 and entorhinal cortex (EC) (functioning as a sparse autoencoder) by contrasting local activity states at different phases in the theta cycle. Based on specific neural data and a recent abstract computational model, we propose a new model called Theremin (Total Hippocampal ERror MINimization) that extends error-driven learning to area CA3 — the mnemonic heart of the hippocampal system. In the model, CA3 responds to the EC monosynaptic input prior to the EC disynaptic input through dentate gyrus (DG), giving rise to a temporal difference between these two activation states, which drives error-driven learning in the EC→CA3 and CA3↔CA3 projections. In effect, DG serves as a teacher to CA3, correcting its patterns into more pattern-separated ones, thereby reducing interference. Results showed that Theremin, compared with our original model, has significantly increased capacity and learning speed. The model makes several novel predictions that can be tested in future studies.

Representations of complex contexts: A role for hippocampus

“…Evidence in humans generally corroborates the research in animal models, suggesting that, as a whole, the HC supports episodic memory by binding together information about items and the context in which they were encountered (Davachi, 2006; Eacott & Gaffan, 2005; Eichenbaum et al, 2007; O’Keefe & Nadel, 1978; Ranganath, 2010; Y. Zheng et al, 2021).…”

“…The hippocampus is a prime example of how form can influence function. Computational models (Marr, 1971; for a non-Hebbian instantiation, see Y. Zheng et al, 2021) propose that the combination of sparse coding in DG and the dense recurrent collaterals that make up the majority of the inputs to CA3 (Amaral & Witter, 1989) enable the hippocampus to retrieve or recover a memory trace given a noisy or incomplete retrieval cue (i.e., “pattern completion” (Marr, 1971; O’Reilly & McClelland, 1994; Yassa & Stark, 2011).…”

“…Zheng et al, 2021) propose that the combination of sparse coding in DG and the dense recurrent collaterals that make up the majority of the inputs to CA3 (Amaral & Witter, 1989) enable the hippocampus to retrieve or recover a memory trace given a noisy or incomplete retrieval cue (i.e., “pattern completion” (Marr, 1971; O’Reilly & McClelland, 1994; Yassa & Stark, 2011). A second factor that is key to understanding hippocampal function is that CA3 and DG are critical sites for binding of information about items and contexts (Knierim et al, 2006; Y. Zheng et al, 2021).…”

Proximity to Boundaries Reveals Spatial Context Representation in Human Hippocampal CA1

“…Computational models propose that the unique anatomical properties of dentate gyrus and CA3 make these subfields ideal for pattern separation, the orthogonalization of highly similar inputs though sparse firing, whereas the CA1 and subiculum subfields are more suited to pattern completion, i.e. preserving features that are common to different inputs (Guzowski et al, 2004;Marr, 1971;Norman and O'Reilly, 2003;Schapiro et al, 2017;Treves and Rolls, 1994;Yassa and Stark, 2011;Y. Zheng et al, 2021).…”