Project Overview
The final project will consist of a small research project on a topic related to computational neuroscience and the application of AI to brain science. You will conduct either original data analysis/modeling work or provide an in-depth exposition of a related topic, evaluated based on final presentation.
Project Structure
- Groups: 1-5 people
- Project presentation: Final weeks of class (time depends on total number of projects)
- Key principle: Keep your scope manageable - not too broad, not overly complicated!
Project Types
Type A: Data Analysis and Modeling
Description: Choose a computational neuroscience or NeuroAI topic that are based on or extend our class material. Conduct original modeling work, simulations, or data analysis.
Example Topics (based on course tutorials):
- Build upon GLM, CNN, RNN, or network analysis tutorials with new applications
- Neural coding extensions: Apply GLMs or encoding models to datasets (e.g., Allen Brain Observatory, IBL data, or other publicly available neural data)
- Representation analysis: Use dimensionality reduction or representational similarity analysis on neural or artificial neural network data
- NeuroAI comparisons: Compare representations between brain areas and artificial neural networks (e.g., using CNN or other models)
- Dynamical systems modeling: Implement state-space models, HMMs, or RNN dynamics for neural data analysis
- Brain network analysis: Apply graph theory methods to connectome data or functional connectivity
- Generative modeling: Use generative models to understand neural population activity or sensory processing
Suggested Approach:
- Start with tutorial provided in class or an open GitHub repo with analysis or modeling code
- Implement a core component of the model/analysis
- Apply it to relevant data or extend it in a novel direction
- Generate insights through systematic exploration of model behavior or data patterns
Type B: Research Summary
Description: Provide an in-depth, didactic explanation of a computational neuroscience or NeuroAI topic that extends beyond our class coverage (no modeling is necessary).
Example Topics:
- Advanced neural coding theories: Information bottleneck, predictive coding, or Bayesian brain hypotheses
- Cutting-edge NeuroAI methods: Foundation models for neuroscience, neural latent models, or brain-inspired AI architectures
- Theory deep dives: Efficient coding theory, criticality in neural networks, or continual learning in biological systems
- Methodological expositions: Advanced dimensionality reduction techniques, causal inference in neuroscience, or modern experimental methods (optogenetics, calcium imaging analysis)
- Cross-disciplinary connections: Quantum approaches to consciousness, information geometry in neural coding, or developmental neuroscience modeling
- AI agent for neuroscience
Suggested Approach:
- Create explanatory diagrams where helpful
- Break down mathematical concepts step-by-step
- Provide intuitive explanations alongside formal descriptions
- Connect the topic to themes from our class
Data Resources
You are encouraged to use real neural data for modeling projects:
Allen Brain Observatory
Visual cortex recordings during behavioral tasks
International Brain Laboratory (IBL)
Standardized decision-making task data across labs
Primate Datasets
Available through various repositories (see Week 2 tutorial materials)
Public Repositories
CRCNS, DANDI, or other open neuroscience datasets