Conference on Cognitive Computational Neuroscience

The hippocampus is important in the rapid learning of 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 beyond what is needed to achieve effective retrieval, causing interference and resulting in a lower learning capacity. Our previous models have utilized a biologically plausible form of error-driven learning in hippocampal CA1 and entorhinal cortex (EC) by contrasting local activity states at different phases in the theta cycle. Based on neural data and a recent model, we propose an updated model Theremin (Total Hippocampal ERror MINimization) that extends error-driven learning to CA3. In Theremin, 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 has significantly increased capacity and learning speed. Theremin could explain learning and memory phenomena like testing effect, and makes novel neurophysiological predictions.