Cortical Column Simulation Using SNNs

Computational Neuroscience Research Lab (CNRL)
University of Tehran

This repository implements a biologically plausible Cortical Column using Spiking Neural Networks (SNNs), inspired by the Thousand Brains Theory of Intelligence by Jeff Hawkins. The project was developed using CoNeX, a framework designed by CNRL for building SNNs based on PyTorch.

---

Overview

Cortical Column Architecture

The simulated Cortical Column (CC) consists of three main components:

1. Layer 56 (L56): Encodes location information as Reference Frames.
2. Layer 4 and Layer 23 (L423): Encodes sensory input.
3. Fully Connected Layer (FC): Facilitates decision-making by integrating processed information.

Key Features

• Biologically plausible design of the Cortical Column.
• Hierarchical structure with specialized roles for each layer.
• Built using CoNeX, a custom framework for SNN development in PyTorch.
• Modular design for easy experimentation and extensibility.

---

Repository Structure

├── /src  
│   ├── /L423            # Layer 4 and Layer 23: Sensory input encoding  
│   │   ├── /synapse     # Synapse-related definitions and implementations  
│   │   ├── /neurons     # Neuron models for L423  
│   │   ├── /network     # Layer-specific network configuration  
│   │   ├── /tools       # Utility functions and tools for L423  
│   │   ├── run_model.py # Script to run the L423 model  
│   │   ├── test_L423.ipynb # Jupyter Notebook for testing  
│   ├── /L56             # Layer 56: Location encoding  
│   │   ├── /synapse  
│   │   ├── /neurons  
│   │   ├── /network  
│   │   ├── /tools  
│   │   ├── run_model.py  
│   │   ├── test_L56.ipynb  
│   ├── /FC              # Fully Connected layer: Decision-making  
│   │   ├── /synapse  
│   │   ├── /neurons  
│   │   ├── /network  
│   │   ├── /tools  
│   │   ├── run_model.py  
│   │   ├── test_FC.ipynb  
├── README.md            # Project documentation  

---

Installation

1. Clone the repository:

   git clone https://github.com/a2l5t8/CNRL-Cortical-Column.git
   cd CNRL-Cortical-Column

2. Install dependencies:

   pip install -r requirements.txt

3. Install CoNeX framework (refer to its documentation).

---

Usage

1. Test the model:
   Navigate to the relevant layer folder (e.g., /L423) and run the corresponding .ipynb file to test the model's behavior.

2. Run the model:
   Use the run_model.py script in each layer folder to execute the simulations. For example:

   python src/L423/run_model.py

3. Modify behaviors:
   Each layer contains synapse, neurons, network, and tools folders, allowing modular customization of behaviors.

---

Contributions

This project is an ongoing effort by the Computational Neuroscience Research Lab (CNRL) at the University of Tehran. Contributions are welcome!

• Submit bug reports and feature requests via the Issues page.
• Fork the repository and create pull requests for enhancements or fixes.

---

References

• Jeff Hawkins - A Thousand Brains: A New Theory of Intelligence
• CoNeX Framework - CoNeX GitHub Repository

---

Happy Coding!