From Egocentric Sensation to Allocentric Space: Adapted Hebbian Learning Model

Overview

The Background Experimentalists proposed that allocentric boundary representations can be generated by pooling across these conjunctive cells — a conjunction of head-direction (HD) and egocentric bounday cells (EBC), effectively rotating egocentric signals into allocentric space.

The Question Motivated by the evidence that the egocentric-to-allocentric transformation may be learned rather than prewired, we ask whether a biologically plausible, local learning rule can acquire it.

The Hypothesis The wall identity occupies a low-dimensional, redundant subspace of conjunctive egocentric boundary boundary cells × head-direction cells. Addtionally, the co-firing of this conjunctive and output allocentric cells makes it ideal for Hebbian-like learning rules. However, standard Hebbian rules fail under the sparsity of conjunctive firing: brief co-activations are too short to consolidate, while long silences let noise erase what was learned.