Grid Cells Are Ubiquitous in Neural Networks
This work explores the ubiquity of grid cells in neural networks, potentially advancing understanding of cognitive mechanisms in AI, but it appears incremental as it builds on prior observations without major breakthroughs.
The study demonstrated that grid cells, important for spatial and non-spatial cognition, can be replicated in deep neural networks for navigation tasks and arise in feedforward networks for non-spatial tasks, supporting grid coding as an effective representation in both biological and artificial networks.
Grid cells are believed to play an important role in both spatial and non-spatial cognition tasks. A recent study observed the emergence of grid cells in an LSTM for path integration. The connection between biological and artificial neural networks underlying the seemingly similarity, as well as the application domain of grid cells in deep neural networks (DNNs), expect further exploration. This work demonstrated that grid cells could be replicated in either pure vision based or vision guided path integration DNNs for navigation under a proper setting of training parameters. We also show that grid-like behaviors arise in feedforward DNNs for non-spatial tasks. Our findings support that the grid coding is an effective representation for both biological and artificial networks.