You are here: Home › Teaching

Teaching

Our teaching covers various aspects of the neurosciences and is part of the MSc Neuroscience program at the University of Freiburg and of the curriculum of the Bachelor and Master of Science Biology offered by the Faculty of Biology.

Search for current lectures in HISinOne.

 

Master thesis projects / Bachelor thesis projects / research projects

 

Velocity representation in medial entorhinal cortex

This is a theoretical project in which the student will extend an existing model of medial entorhinal grid cells to account for theta phase precession of spiking neurons. The simulated firing patterns will be compared to openly available experimental data to test the hypothesis that the grid cell population represents translations in 2-dimensional space. The hypothesis will be evaluated using methods from machine learning.

The project is suitable for a master thesis and a research project. It can also be Bachelor thesis work if students have good programming skills.

Contact:

 

Learning-Induced Restructuring of Neural Manifolds in the Prefrontal Cortex

Neural population activity in the medial prefrontal cortex (mPFC) evolves during learning, but whether task representations emerge de novo or build upon preexisting neural schemas remains unresolved. While hippocampal CA1 undergoes global remapping, PFC representations are thought to transition smoothly within a stable, low-dimensional manifold, supporting task generalization rather than environmental remapping. This project will apply manifold learning techniques to examine how place-selective and task-selective cells reconfigure their population dynamics pre- and post-learning, determining whether learning compresses existing dimensions (schema-based learning) or expands into new representational states (de novo learning).

To quantify these changes, we will construct latent neural manifolds and apply manifold alignment metrics to measure shifts in representational geometry. This approach will reveal whether post-learning activity remains within the pre-learning structure or forms a novel representational subspace. By providing a geometric framework for neural reorganization, this project will clarify how cortical circuits balance stability and plasticity during learning.

Contact:

 

Hippocampus-Inspired Deep Reinforcement Learning Agent for Egocentric Visual Navigation

Project Overview:
This NeuroAI project bridges neuroscience and cutting-edge AI by leveraging insights from the hippocampus—the brain’s navigation center—to inspire advanced deep reinforcement learning (RL) algorithms. Instead of traditional approaches, we adopt a physiology-centered perspective to study hippocampal function. By embedding the known anatomical and computational structure of the hippocampus into a deep RL agent, we aim to examine prevalent theories and provide a deeper, principled explanation of spatial representation in the hippocampus.

This dual approach not only advances our understanding of neural processes underlying navigation but also paves the way for designing more biologically grounded NeuroAI systems.

Key Objectives:
Develop Hippocampus-inspired NeuroAI Architectures:
    • Design and refine deep RL agents using hippocampal anatomy, namely the sequential recurrent connections in the CA3 region and sparse projection from dentate gyrus to CA3. #reservoir_computing
    • Integrate modern best-performing architectures like Mamba and DreamerV3 with neuro-inspired principles and explore the minimalist architectures that is efficient and allow implicit world modeling.
    Create and Customize Virtual Environments:
      • Deploy and evaluate RL agents in diverse simulated environments (e.g., Webots, DeepMind Lab, CARLA, Habitat, VizDOOM, MiniWorld, Madrona, JaxMARL).
      • Customize these environments for specialized navigation and neuroscience tasks that involves high-level cognition.
      • Explore novel benchmarks beyond navigation in physical space, such as treating text generation as navigation within text embedding spaces (like DeepSeek-R1) and using LLMs as environmental simulators / reward models.
      Analyze and Reverse Engineer Neural Representations:
        • Investigate how the agent’s neural network encodes spatial information and supports navigation and planning.
        • Compare hidden unit behaviors with phenomena in neurophysiological recordings to reveal parallels between artificial and biological navigation systems.

         

        Why Join This Project?

        Interdisciplinary Experience: Merge theoretical neuroscience with state-of-the-art AI research in an inspiring NeuroAI setting.

        Technical Growth: Enhance your skills using modern tools like PyTorch and Gymnasium, alongside game engines and simulation platforms.

        Strong Theoretical & Collaborative Environment: Work alongside physicists and neuroscientists in a lab with a robust theoretical foundation, offering the perfect bridge between neuroscience and AI, all within a friendly, supportive, and collaborative atmosphere.

        Customization and Flexibility: Tailor your involvement based on your programming, machine learning, and scientific interests—from architecture design to benchmark development.

        Fun and Impactful Work: Enjoy a dynamic, game-like environment while contributing to transformative research that redefines the future of NeuroAI.

        Contact: For more details and to join our exciting journey, please reach out at