Exponential Map Models as an Interpretable Framework for Generating Neural Spatial Representations

Matrix-exponential generator models map physical locations to neural population vectors without gradient optimization, and algebraic constraints yield exact path integration, equinorm codes, and periodic grid fields.

• Why it matters: Neuroscience and AI lack transparent ways to tell whether learned spatial representations reflect general principles or architecture-specific artifacts, and whether they generalize beyond training data.
• What they did: They build a first-principles framework where generator matrices produce neural vectors via the matrix exponential; commuting generators give trajectory-invariant exact path integration, skew-symmetry gives equinorm translation-invariant similarities, and metric preservation constrains spectra to balanced wavevector/root-of-unity configurations.
• The result: The same algebraic setup produces localized place-like fields and context-dependent remapping, fits recorded grid and place cell data, extrapolates rate maps beyond the fitted region, and enables testable predictions for entorhinal circuits.

AI summary of the abstract · may contain errors · not medical advice · read the paper.