Abstract
The hippocampal-entorhinal system is important for spatial and relational memory tasks. We formally link these domains; provide a mechanistic understanding of the hippocampal role in generalisation; and offer unifying principles underlying many entorhinal and hippocampal cell-types. We propose medial entorhinal cells form a basis describing structural knowledge, and hippocampal cells link this basis with sensory representations. Adopting these principles, we introduce the Tolman-Eichenbaum machine (TEM). After learning, TEM entorhinal cells include grid, band, border and object-vector cells. Hippocampal cells include place and landmark cells, remapping between environments. Crucially, TEM also predicts empirically recorded representations in complex non-spatial tasks. TEM predicts hippocampal remapping is not random as previously believed. Rather structural knowledge is preserved across environments. We confirm this in simultaneously recorded place and grid cells.
One Sentence Summary Simple principles of representation and generalisation unify spatial and non-spatial accounts of hippocampus and explain many cell representations.
Footnotes
↵† Joint first author
Removed some typos.
1
is an intermediary variable retrieved via the memory Mt−1 from xt - i.e. it represents x≤t and Mt−1 in the posterior for gt.
