Gridness-score analysis
Techniqueintermediate
Computes a spatial autocorrelogram of each unit's activity as a function of 2D physical position, then scores hexagonal periodicity (gridness) by comparing the autocorrelogram's correlation at 60 degree/120 degree rotations against 30 degree/90 degree/150 degree rotations — the standard rodent-electrophysiology metric (Sargolini et al.) applied to artificial-network unit activity.
Used in (2 observations)
structure: Torus · models: LSTM path-integration module + A3C navigation agent · paper: Vector-based Navigation Using Grid-like Representations in Artificial Agents
structure: Torus · models: Tolman-Eichenbaum Machine (trained on spatial + non-spatial relational structure-learning tasks) · paper: The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation