Modeling Hippocampal Spatial Navigation and Memory using Brain-Inspired AI

Delve into the fascinating intersection of neuroscience and AI by creating brain-inspired AI models of the hippocampus to investigate how we navigate our world and form new memories.

The hippocampus is a crucial brain region known for its fundamental role in spatial navigation and the formation of new memories. This is achieved through specialized neurons like place cells, which fire at specific locations, and grid cells, which represent space in a coordinate-like system. While we know these cells are fundamental, the precise computational principles that govern their interactions to enable spatial navigation and memory remain an open and exciting question. In this project, you will bridge insights from neuroscience with AI algorithms to develop brain-inspired AI models to investigate how neural representations of space can be learned and used for navigation tasks. By creating more biologically plausible models, this project aims to bring new insights about how the hippocampus learns to represent and navigate the world.

Goal

The topic gives room for multiple individual or collaborative projects tailored to the background and interests of the student(s). The primary goals are to:

• Develop and train novel machine learning models (e.g., using reinforcement learning, generative models, or neural networks) that incorporate biologically plausible constraints from the hippocampus.
• Investigate how these models build "cognitive maps" and form spatial representations of their environment, comparing the learned representations to real neural data.
• Investigate how artificial agents can use cognitive maps to perform flexible, goal-directed navigation, such as finding novel shortcuts, and compare their emergent behaviors to experimental observations.
• Use the models to generate testable predictions about the neural mechanisms underlying spatial learning and memory-guided decision-making.
• Explore how principles from neuroscience can lead to more efficient and generalizable navigation algorithms in artificial agents.

Learning outcome

Upon completion of this project, you will have:

• Deep insights into the structure and function of the hippocampus and its role in spatial cognition.
• A strong theoretical and practical understanding of advanced machine learning algorithms and how they can model complex biological systems.
• Hands-on experience with modern machine learning frameworks such as PyTorch or Jax.
• Valuable experience in interdisciplinary research, learning to integrate and communicate concepts from both neuroscience and computer science.

Qualifications

Most importantly, we are looking for curious, enthusiastic, and goal-oriented students who are excited to work at the intersection of two dynamic fields.

Need-to-have:

• A solid background in at least one of the following: machine learning, computer science, physics, or applied mathematics.
• A good understanding of core mathematical concepts: linear algebra, calculus, and probability theory.
• Strong programming skills, particularly in Python.

Nice-to-have:

• Previous experience with machine learning frameworks like PyTorch, TensorFlow, or Jax.
• A demonstrated interest in neuroscience, cognitive science, or biology.
• Familiarity with machine learning concepts.

Supervisors

• Mikkel Elle Lepperød
• Nicolai Haug

Collaboration partners

• University of Oslo

References

• The master's thesis is associated with the research project bioAI – Biologically inspired Artificial Intelligence