The Role of Grid Cells in Spatial Cognition

Project Overview

Cognitive neuroscience can help provide mechanistic explanations of how human cognition could work. At the same time, it can inspire computational models in artificial intelligence (e.g., for robots with spatial abilities).

Study the literature related to grid cells in the brain, from the computational perspective.

Using [1], implement a small neural network model containing grid cells. Train, test, and analyze the model on a small selected task involving spatial representations.

References

• [1] Whittington, James C.R. et al. (2020) The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation. Cell, 183(5), 1249-1263
• [2] Ling L. Dong L.L., Fiete R. (2024) Grid Cells in Cognition: Mechanisms and Function. Annual Review of Neuroscience
• [3] Hawkins J. et al. (2019) A framework for intelligence and cortical function based on grid cells in the neocortex. Frontiers in Neural Circuits

Work on the project

Introduction

    In this project, we will be studying a paper authored by James Whittington called "The Tolman-Eichenbaum Machine: Unifying Space and Relational Memory through Generalization in the Hippocampal Formation" [1]. We will work towards understanding the high-level concepts proposed in this paper, learning high-level concepts of cognitive science as preliminaries and learning about neural network architectures used in this paper.

    Once we have sufficient preliminary knowledge to replicate the paper's findings, we will move on to implementing Tolman-Eichenbaum Machine (TEM) using the methods proposed in this paper as guides. From this we will then explore environments/learning methods on our TEM which the authors haven't explored and will attempt to draw conclusions and explore the role of grid cells in spatial cognition of the TEM.

    In order to do any work on this project, we must first familiarize ourselves with the principles and methods of cognitive science as well get ahead of our study program's curriculum to be able to work with proposed neural network models and methods in [1]. Since we are implementing the same TEM as proposed and thoroughly described in this paper, we will not dedicate as much time to implementing it in code. Instead, we will focus on research aspects and try and further evaluate the possibilities of TEM as well as explore possible inclusion of the TEM in other, cognitive task-related projects.

Timeline

• 20. 10. 2025 - START OF THE PROJECT.

  First meeting with guiding teacher, expressing interest in CogSci, picking the right topic (not too advanced, but integrated enough to classify as year project)

• 24. 10. 2025 - ESTABLISHED TOPIC, START OF RESEARCH.

  First readings of the paper. Filling out gaps in knowledge with the help of guiding teacher. Gathering resources and studying.

• 10. 11. 2025 - ESTABLISHED PROJECT WEBSITE.

  Creating a HTML website supposed to provide insight into the project's status, direction and work.

• 17. 11. 2025 - CONTINUED RESEARCH, MORE MEETINGS.

  Gaining confidence in the subject, frequent meetings with the guiding teacher in order to stay on the right track. Looking into more technical aspect of the project, planning development phase.

• 23. 1. 2026 - PROJECT HALF-YEAR REPORT.

  [ PROJECT REPORT ]